Datasets:

IAMJB

/

EMNLP2023-papers

like 0

https://aclanthology.org/2023.emnlp-main.701.bib

https://aclanthology.org/2023.emnlp-main.701/

@inproceedings{yerukola-etal-2023-dont, title = "Don{'}t Take This Out of Context!: On the Need for Contextual Models and Evaluations for Stylistic Rewriting", author = "Yerukola, Akhila and Zhou, Xuhui and Clark, Elizabeth and Sap, Maarten", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.701", doi = "10.18653/v1/2023.emnlp-main.701", pages = "11419--11444", abstract = "Most existing stylistic text rewriting methods and evaluation metrics operate on a sentence level, but ignoring the broader context of the text can lead to preferring generic, ambiguous, and incoherent rewrites. In this paper, we investigate integrating the preceding textual context into both the $\textit{rewriting}$ and $\textit{evaluation}$ stages of stylistic text rewriting, and introduce a new composite contextual evaluation metric $\texttt{CtxSimFit}$ that combines similarity to the original sentence with contextual cohesiveness. We comparatively evaluate non-contextual and contextual rewrites in formality, toxicity, and sentiment transfer tasks. Our experiments show that humans significantly prefer contextual rewrites as more fitting and natural over non-contextual ones, yet existing sentence-level automatic metrics (e.g., ROUGE, SBERT) correlate poorly with human preferences ($\rho$=0{--}0.3). In contrast, human preferences are much better reflected by both our novel $\texttt{CtxSimFit}$ ($\rho$=0.7{--}0.9) as well as proposed context-infused versions of common metrics ($\rho$=0.4{--}0.7). Overall, our findings highlight the importance of integrating context into the generation and especially the evaluation stages of stylistic text rewriting.", }

Most existing stylistic text rewriting methods and evaluation metrics operate on a sentence level, but ignoring the broader context of the text can lead to preferring generic, ambiguous, and incoherent rewrites. In this paper, we investigate integrating the preceding textual context into both the $\textit{rewriting}$ and $\textit{evaluation}$ stages of stylistic text rewriting, and introduce a new composite contextual evaluation metric $\texttt{CtxSimFit}$ that combines similarity to the original sentence with contextual cohesiveness. We comparatively evaluate non-contextual and contextual rewrites in formality, toxicity, and sentiment transfer tasks. Our experiments show that humans significantly prefer contextual rewrites as more fitting and natural over non-contextual ones, yet existing sentence-level automatic metrics (e.g., ROUGE, SBERT) correlate poorly with human preferences ($\rho$=0{--}0.3). In contrast, human preferences are much better reflected by both our novel $\texttt{CtxSimFit}$ ($\rho$=0.7{--}0.9) as well as proposed context-infused versions of common metrics ($\rho$=0.4{--}0.7). Overall, our findings highlight the importance of integrating context into the generation and especially the evaluation stages of stylistic text rewriting.

Don't Take This Out of Context!: On the Need for Contextual Models and Evaluations for Stylistic Rewriting

emnlp-main.701

Poster

https://aclanthology.org/2023.emnlp-main.702.bib

https://aclanthology.org/2023.emnlp-main.702/

@inproceedings{rosset-etal-2023-axiomatic, title = "Axiomatic Preference Modeling for Longform Question Answering", author = "Rosset, Corby and Zheng, Guoqing and Dibia, Victor and Awadallah, Ahmed and Bennett, Paul", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.702", doi = "10.18653/v1/2023.emnlp-main.702", pages = "11445--11475", abstract = "The remarkable abilities of large language models (LLMs) like ChatGPT and GPT-4 partially stem from the post-training processes involving human preferences encoded within a reward model as part of a Reinforcement Learning from Human Feedback (RLHF) regimen. These reward models (RMs) often lack direct knowledge of why, or under what principles, the preferences annotations were made. In this study, we identify principles that guide RMs to better align with human preferences, and then develop an axiomatic framework to generate a rich variety of preference signals to uphold them. We use these axiomatic signals to train a model for the scoring answers to longform questions. Our approach yields a \textbf{Preference Model} with only about 220M parameters that agrees with gold human-annotated preference labels more often than GPT-4. The contributions of this work include: training a standalone preference model that can score human- and LLM-generated answers on the same scale; developing an axiomatic framework for generating training data pairs tailored to certain principles; and showing that a small amount of axiomatic signals can help small models outperform GPT-4 in preference scoring. We intend to release our axiomatic data and model.", }

The remarkable abilities of large language models (LLMs) like ChatGPT and GPT-4 partially stem from the post-training processes involving human preferences encoded within a reward model as part of a Reinforcement Learning from Human Feedback (RLHF) regimen. These reward models (RMs) often lack direct knowledge of why, or under what principles, the preferences annotations were made. In this study, we identify principles that guide RMs to better align with human preferences, and then develop an axiomatic framework to generate a rich variety of preference signals to uphold them. We use these axiomatic signals to train a model for the scoring answers to longform questions. Our approach yields a \textbf{Preference Model} with only about 220M parameters that agrees with gold human-annotated preference labels more often than GPT-4. The contributions of this work include: training a standalone preference model that can score human- and LLM-generated answers on the same scale; developing an axiomatic framework for generating training data pairs tailored to certain principles; and showing that a small amount of axiomatic signals can help small models outperform GPT-4 in preference scoring. We intend to release our axiomatic data and model.

Axiomatic Preference Modeling for Longform Question Answering

emnlp-main.702

2312.02206

https://huggingface.co/papers/2312.02206

Poster

https://aclanthology.org/2023.emnlp-main.703.bib

https://aclanthology.org/2023.emnlp-main.703/

@inproceedings{russo-etal-2023-countering, title = "Countering Misinformation via Emotional Response Generation", author = "Russo, Daniel and Kaszefski-Yaschuk, Shane and Staiano, Jacopo and Guerini, Marco", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.703", doi = "10.18653/v1/2023.emnlp-main.703", pages = "11476--11492", abstract = "The proliferation of misinformation on social media platforms (SMPs) poses a significant danger to public health, social cohesion and ultimately democracy. Previous research has shown how social correction can be an effective way to curb misinformation, by engaging directly in a constructive dialogue with users who spread {--} often in good faith {--} misleading messages. Although professional fact-checkers are crucial to debunking viral claims, they usually do not engage in conversations on social media. Thereby, significant effort has been made to automate the use of fact-checker material in social correction; however, no previous work has tried to integrate it with the style and pragmatics that are commonly employed in social media communication. To fill this gap, we present VerMouth, the first large-scale dataset comprising roughly 12 thousand claim-response pairs (linked to debunking articles), accounting for both SMP-style and basic emotions, two factors which have a significant role in misinformation credibility and spreading. To collect this dataset we used a technique based on an author-reviewer pipeline, which efficiently combines LLMs and human annotators to obtain high-quality data. We also provide comprehensive experiments showing how models trained on our proposed dataset have significant improvements in terms of output quality and generalization capabilities.", }

The proliferation of misinformation on social media platforms (SMPs) poses a significant danger to public health, social cohesion and ultimately democracy. Previous research has shown how social correction can be an effective way to curb misinformation, by engaging directly in a constructive dialogue with users who spread {--} often in good faith {--} misleading messages. Although professional fact-checkers are crucial to debunking viral claims, they usually do not engage in conversations on social media. Thereby, significant effort has been made to automate the use of fact-checker material in social correction; however, no previous work has tried to integrate it with the style and pragmatics that are commonly employed in social media communication. To fill this gap, we present VerMouth, the first large-scale dataset comprising roughly 12 thousand claim-response pairs (linked to debunking articles), accounting for both SMP-style and basic emotions, two factors which have a significant role in misinformation credibility and spreading. To collect this dataset we used a technique based on an author-reviewer pipeline, which efficiently combines LLMs and human annotators to obtain high-quality data. We also provide comprehensive experiments showing how models trained on our proposed dataset have significant improvements in terms of output quality and generalization capabilities.

Countering Misinformation via Emotional Response Generation

emnlp-main.703

2311.10587

Poster

https://aclanthology.org/2023.emnlp-main.704.bib

https://aclanthology.org/2023.emnlp-main.704/

@inproceedings{zhang-etal-2023-seq2seq, title = "Seq2seq is All You Need for Coreference Resolution", author = "Zhang, Wenzheng and Wiseman, Sam and Stratos, Karl", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.704", doi = "10.18653/v1/2023.emnlp-main.704", pages = "11493--11504", abstract = "Existing works on coreference resolution suggest that task-specific models are necessary to achieve state-of-the-art performance. In this work, we present compelling evidence that such models are not necessary. We finetune a pretrained seq2seq transformer to map an input document to a tagged sequence encoding the coreference annotation. Despite the extreme simplicity, our model outperforms or closely matches the best coreference systems in the literature on an array of datasets. We consider an even simpler version of seq2seq that generates only the tagged spans and find it highly performant. Our analysis shows that the model size, the amount of supervision, and the choice of sequence representations are key factors in performance.", }

Existing works on coreference resolution suggest that task-specific models are necessary to achieve state-of-the-art performance. In this work, we present compelling evidence that such models are not necessary. We finetune a pretrained seq2seq transformer to map an input document to a tagged sequence encoding the coreference annotation. Despite the extreme simplicity, our model outperforms or closely matches the best coreference systems in the literature on an array of datasets. We consider an even simpler version of seq2seq that generates only the tagged spans and find it highly performant. Our analysis shows that the model size, the amount of supervision, and the choice of sequence representations are key factors in performance.

Seq2seq is All You Need for Coreference Resolution

emnlp-main.704

2310.13774

Oral

https://aclanthology.org/2023.emnlp-main.705.bib

https://aclanthology.org/2023.emnlp-main.705/

@inproceedings{fucci-etal-2023-integrating, title = "Integrating Language Models into Direct Speech Translation: An Inference-Time Solution to Control Gender Inflection", author = "Fucci, Dennis and Gaido, Marco and Papi, Sara and Cettolo, Mauro and Negri, Matteo and Bentivogli, Luisa", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.705", doi = "10.18653/v1/2023.emnlp-main.705", pages = "11505--11517", abstract = "When translating words referring to the speaker, speech translation (ST) systems should not resort to default masculine generics nor rely on potentially misleading vocal traits. Rather, they should assign gender according to the speakers{'} preference. The existing solutions to do so, though effective, are hardly feasible in practice as they involve dedicated model re-training on gender-labeled ST data. To overcome these limitations, we propose the first inference-time solution to control speaker-related gender inflections in ST. Our approach partially replaces the (biased) internal language model (LM) implicitly learned by the ST decoder with gender-specific external LMs. Experiments on en$\rightarrow$es/fr/it show that our solution outperforms the base models and the best training-time mitigation strategy by up to 31.0 and 1.6 points in gender accuracy, respectively, for feminine forms. The gains are even larger (up to 32.0 and 3.4) in the challenging condition where speakers{'} vocal traits conflict with their gender.", }

When translating words referring to the speaker, speech translation (ST) systems should not resort to default masculine generics nor rely on potentially misleading vocal traits. Rather, they should assign gender according to the speakers{'} preference. The existing solutions to do so, though effective, are hardly feasible in practice as they involve dedicated model re-training on gender-labeled ST data. To overcome these limitations, we propose the first inference-time solution to control speaker-related gender inflections in ST. Our approach partially replaces the (biased) internal language model (LM) implicitly learned by the ST decoder with gender-specific external LMs. Experiments on en$\rightarrow$es/fr/it show that our solution outperforms the base models and the best training-time mitigation strategy by up to 31.0 and 1.6 points in gender accuracy, respectively, for feminine forms. The gains are even larger (up to 32.0 and 3.4) in the challenging condition where speakers{'} vocal traits conflict with their gender.

Integrating Language Models into Direct Speech Translation: An Inference-Time Solution to Control Gender Inflection

emnlp-main.705

2310.15752

https://huggingface.co/papers/2310.15752

Oral

https://aclanthology.org/2023.emnlp-main.706.bib

https://aclanthology.org/2023.emnlp-main.706/

@inproceedings{jiayang-etal-2023-storyanalogy, title = "{S}tory{A}nalogy: Deriving Story-level Analogies from Large Language Models to Unlock Analogical Understanding", author = "Jiayang, Cheng and Qiu, Lin and Chan, Tsz and Fang, Tianqing and Wang, Weiqi and Chan, Chunkit and Ru, Dongyu and Guo, Qipeng and Zhang, Hongming and Song, Yangqiu and Zhang, Yue and Zhang, Zheng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.706", doi = "10.18653/v1/2023.emnlp-main.706", pages = "11518--11537", abstract = "Analogy-making between narratives is crucial for human reasoning. In this paper, we evaluate the ability to identify and generate analogies by constructing a first-of-its-kind large-scale story-level analogy corpus, StoryAnalogy, which contains 24K story pairs from diverse domains with human annotations on two similarities from the extended Structure-Mapping Theory. We design a set of tests on StoryAnalogy, presenting the first evaluation of story-level analogy identification and generation. Interestingly, we find that the analogy identification tasks are incredibly difficult not only for sentence embedding models but also for the recent large language models (LLMs) such as ChatGPT and LLaMa. ChatGPT, for example, only achieved around 30{\%} accuracy in multiple-choice questions (compared to over 85{\%} accuracy for humans). Furthermore, we observe that the data in StoryAnalogy can improve the quality of analogy generation in LLMs, where a fine-tuned FlanT5-xxl model achieves comparable performance to zero-shot ChatGPT.", }

Analogy-making between narratives is crucial for human reasoning. In this paper, we evaluate the ability to identify and generate analogies by constructing a first-of-its-kind large-scale story-level analogy corpus, StoryAnalogy, which contains 24K story pairs from diverse domains with human annotations on two similarities from the extended Structure-Mapping Theory. We design a set of tests on StoryAnalogy, presenting the first evaluation of story-level analogy identification and generation. Interestingly, we find that the analogy identification tasks are incredibly difficult not only for sentence embedding models but also for the recent large language models (LLMs) such as ChatGPT and LLaMa. ChatGPT, for example, only achieved around 30{\%} accuracy in multiple-choice questions (compared to over 85{\%} accuracy for humans). Furthermore, we observe that the data in StoryAnalogy can improve the quality of analogy generation in LLMs, where a fine-tuned FlanT5-xxl model achieves comparable performance to zero-shot ChatGPT.

StoryAnalogy: Deriving Story-level Analogies from Large Language Models to Unlock Analogical Understanding

emnlp-main.706

2310.12874

https://huggingface.co/papers/2310.12874

Poster

https://aclanthology.org/2023.emnlp-main.707.bib

https://aclanthology.org/2023.emnlp-main.707/

@inproceedings{chang-etal-2023-beyond, title = "Beyond Detection: A Defend-and-Summarize Strategy for Robust and Interpretable Rumor Analysis on Social Media", author = "Chang, Yi-Ting and Song, Yun-Zhu and Chen, Yi-Syuan and Shuai, Hong-Han", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.707", doi = "10.18653/v1/2023.emnlp-main.707", pages = "11538--11556", abstract = "As the impact of social media gradually escalates, people are more likely to be exposed to indistinguishable fake news. Therefore, numerous studies have attempted to detect rumors on social media by analyzing the textual content and propagation paths. However, fewer works on rumor detection tasks consider the malicious attacks commonly observed at response level. Moreover, existing detection models have poor interpretability. To address these issues, we propose a novel framework named **D**efend-**A**nd-**S**ummarize (DAS) based on the concept that responses sharing similar opinions should exhibit similar features. Specifically, DAS filters out the attack responses and summarizes the responsive posts of each conversation thread in both extractive and abstractive ways to provide multi-perspective prediction explanations. Furthermore, we enhance our detection architecture with the transformer and Bi-directional Graph Convolutional Networks. Experiments on three public datasets, *i.e.*, RumorEval2019, Twitter15, and Twitter16, demonstrate that our DAS defends against malicious attacks and provides prediction explanations, and the proposed detection model achieves state-of-the-art.", }

As the impact of social media gradually escalates, people are more likely to be exposed to indistinguishable fake news. Therefore, numerous studies have attempted to detect rumors on social media by analyzing the textual content and propagation paths. However, fewer works on rumor detection tasks consider the malicious attacks commonly observed at response level. Moreover, existing detection models have poor interpretability. To address these issues, we propose a novel framework named **D**efend-**A**nd-**S**ummarize (DAS) based on the concept that responses sharing similar opinions should exhibit similar features. Specifically, DAS filters out the attack responses and summarizes the responsive posts of each conversation thread in both extractive and abstractive ways to provide multi-perspective prediction explanations. Furthermore, we enhance our detection architecture with the transformer and Bi-directional Graph Convolutional Networks. Experiments on three public datasets, *i.e.*, RumorEval2019, Twitter15, and Twitter16, demonstrate that our DAS defends against malicious attacks and provides prediction explanations, and the proposed detection model achieves state-of-the-art.

Beyond Detection: A Defend-and-Summarize Strategy for Robust and Interpretable Rumor Analysis on Social Media

emnlp-main.707

Poster

https://aclanthology.org/2023.emnlp-main.708.bib

https://aclanthology.org/2023.emnlp-main.708/

@inproceedings{liu-etal-2023-crystal, title = "Crystal: Introspective Reasoners Reinforced with Self-Feedback", author = "Liu, Jiacheng and Pasunuru, Ramakanth and Hajishirzi, Hannaneh and Choi, Yejin and Celikyilmaz, Asli", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.708", doi = "10.18653/v1/2023.emnlp-main.708", pages = "11557--11572", abstract = "Extensive work has shown that the performance and interpretability of commonsense reasoning can be improved via knowledge-augmented reasoning methods, where the knowledge that underpins the reasoning process is explicitly verbalized and utilized. However, existing implementations, including {``}chain-of-thought{''} and its variants, fall short in capturing the *introspective* nature of knowledge required in commonsense reasoning, and in accounting for the mutual adaptation between the generation and utilization of knowledge. We propose a novel method to develop an introspective commonsense reasoner, **Crystal**. To tackle commonsense problems, it first introspects for knowledge statements related to the given question, and subsequently makes an informed prediction that is grounded in the previously introspected knowledge. The knowledge introspection and knowledge-grounded reasoning modes of the model are tuned via reinforcement learning to mutually adapt, where the reward derives from the feedback given by the model itself. Experiments show that Crystal significantly outperforms both the standard supervised finetuning and chain-of-thought distilled methods, and enhances the transparency of the commonsense reasoning process. Our work ultimately validates the feasibility and potential of reinforcing a neural model with self-feedback.", }

Extensive work has shown that the performance and interpretability of commonsense reasoning can be improved via knowledge-augmented reasoning methods, where the knowledge that underpins the reasoning process is explicitly verbalized and utilized. However, existing implementations, including {``}chain-of-thought{''} and its variants, fall short in capturing the *introspective* nature of knowledge required in commonsense reasoning, and in accounting for the mutual adaptation between the generation and utilization of knowledge. We propose a novel method to develop an introspective commonsense reasoner, **Crystal**. To tackle commonsense problems, it first introspects for knowledge statements related to the given question, and subsequently makes an informed prediction that is grounded in the previously introspected knowledge. The knowledge introspection and knowledge-grounded reasoning modes of the model are tuned via reinforcement learning to mutually adapt, where the reward derives from the feedback given by the model itself. Experiments show that Crystal significantly outperforms both the standard supervised finetuning and chain-of-thought distilled methods, and enhances the transparency of the commonsense reasoning process. Our work ultimately validates the feasibility and potential of reinforcing a neural model with self-feedback.

Crystal: Introspective Reasoners Reinforced with Self-Feedback

emnlp-main.708

2310.04921

https://huggingface.co/papers/2310.04921

Oral

https://aclanthology.org/2023.emnlp-main.709.bib

https://aclanthology.org/2023.emnlp-main.709/

@inproceedings{zhu-etal-2023-diffs2ut, title = "{D}iff{S}2{UT}: A Semantic Preserving Diffusion Model for Textless Direct Speech-to-Speech Translation", author = "Zhu, Yongxin and Gao, Zhujin and Zhou, Xinyuan and Zhongyi, Ye and Xu, Linli", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.709", doi = "10.18653/v1/2023.emnlp-main.709", pages = "11573--11583", abstract = "While Diffusion Generative Models have achieved great success on image generation tasks, how to efficiently and effectively incorporate them into speech generation especially translation tasks remains a non-trivial problem. Specifically, due to the low information density of speech data, the transformed discrete speech unit sequence is much longer than the corresponding text transcription, posing significant challenges to existing auto-regressive models. Furthermore, it is not optimal to brutally apply discrete diffusion on the speech unit sequence while disregarding the continuous space structure, which will degrade the generation performance significantly. In this paper, we propose a novel diffusion model by applying the diffusion forward process in the continuous speech representation space, while employing the diffusion backward process in the discrete speech unit space. In this way, we preserve the semantic structure of the continuous speech representation space in the diffusion process and integrate the continuous and discrete diffusion models. We conduct extensive experiments on the textless direct speech-to-speech translation task, where the proposed method achieves comparable results to the computationally intensive auto-regressive baselines (500 steps on average) with significantly fewer decoding steps (50 steps).", }

While Diffusion Generative Models have achieved great success on image generation tasks, how to efficiently and effectively incorporate them into speech generation especially translation tasks remains a non-trivial problem. Specifically, due to the low information density of speech data, the transformed discrete speech unit sequence is much longer than the corresponding text transcription, posing significant challenges to existing auto-regressive models. Furthermore, it is not optimal to brutally apply discrete diffusion on the speech unit sequence while disregarding the continuous space structure, which will degrade the generation performance significantly. In this paper, we propose a novel diffusion model by applying the diffusion forward process in the continuous speech representation space, while employing the diffusion backward process in the discrete speech unit space. In this way, we preserve the semantic structure of the continuous speech representation space in the diffusion process and integrate the continuous and discrete diffusion models. We conduct extensive experiments on the textless direct speech-to-speech translation task, where the proposed method achieves comparable results to the computationally intensive auto-regressive baselines (500 steps on average) with significantly fewer decoding steps (50 steps).

DiffS2UT: A Semantic Preserving Diffusion Model for Textless Direct Speech-to-Speech Translation

emnlp-main.709

2310.17570

https://huggingface.co/papers/2310.17570

Poster

https://aclanthology.org/2023.emnlp-main.710.bib

https://aclanthology.org/2023.emnlp-main.710/

@inproceedings{sui-etal-2023-biofeg, title = "{B}io{FEG}: Generate Latent Features for Biomedical Entity Linking", author = "Sui, Xuhui and Zhang, Ying and Cai, Xiangrui and Song, Kehui and Zhou, Baohang and Yuan, Xiaojie and Zhang, Wensheng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.710", doi = "10.18653/v1/2023.emnlp-main.710", pages = "11584--11593", abstract = "Biomedical entity linking is an essential task in biomedical text processing, which aims to map entity mentions in biomedical text, such as clinical notes, to standard terms in a given knowledge base. However, this task is challenging due to the rarity of many biomedical entities in real-world scenarios, which often leads to a lack of annotated data for them. Limited by understanding these unseen entities, traditional biomedical entity linking models suffer from multiple types of linking errors. In this paper, we propose a novel latent feature generation framework BioFEG to address these challenges. Specifically, our BioFEG leverages domain knowledge to train a generative adversarial network, which generates latent semantic features of corresponding mentions for unseen entities. Utilizing these features, we fine-tune our entity encoder to capture fine-grained coherence information of unseen entities and better understand them. This allows models to make linking decisions more accurately, particularly for ambiguous mentions involving rare entities. Extensive experiments on the two benchmark datasets demonstrate the superiority of our proposed framework.", }

Biomedical entity linking is an essential task in biomedical text processing, which aims to map entity mentions in biomedical text, such as clinical notes, to standard terms in a given knowledge base. However, this task is challenging due to the rarity of many biomedical entities in real-world scenarios, which often leads to a lack of annotated data for them. Limited by understanding these unseen entities, traditional biomedical entity linking models suffer from multiple types of linking errors. In this paper, we propose a novel latent feature generation framework BioFEG to address these challenges. Specifically, our BioFEG leverages domain knowledge to train a generative adversarial network, which generates latent semantic features of corresponding mentions for unseen entities. Utilizing these features, we fine-tune our entity encoder to capture fine-grained coherence information of unseen entities and better understand them. This allows models to make linking decisions more accurately, particularly for ambiguous mentions involving rare entities. Extensive experiments on the two benchmark datasets demonstrate the superiority of our proposed framework.

BioFEG: Generate Latent Features for Biomedical Entity Linking

emnlp-main.710

Poster

https://aclanthology.org/2023.emnlp-main.711.bib

https://aclanthology.org/2023.emnlp-main.711/

@inproceedings{xiong-etal-2023-trigo, title = "{TRIGO}: Benchmarking Formal Mathematical Proof Reduction for Generative Language Models", author = "Xiong, Jing and Shen, Jianhao and Yuan, Ye and Wang, Haiming and Yin, Yichun and Liu, Zhengying and Li, Lin and Guo, Zhijiang and Cao, Qingxing and Huang, Yinya and Zheng, Chuanyang and Liang, Xiaodan and Zhang, Ming and Liu, Qun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.711", doi = "10.18653/v1/2023.emnlp-main.711", pages = "11594--11632", abstract = "Automated theorem proving (ATP) has become an appealing domain for exploring the reasoning ability of the recent successful generative language models. However, current ATP benchmarks are mainly focus on symbolic inference, but rarely involve the understanding of complex number combination reasoning. In this work, we propose TRIGO, an ATP benchmark that not only requires a model to reduce a trigonometric expression with step-by-step proof but also evaluates a generative LM{'}s reasoning ability on formulas and capability to manipulate, group, and factor number terms. We gather trigonometric expressions and their reduced forms from web, annotate the simplification process manually, and translate it into the {``}Lean{''} formal language system. We then automatically generate additional examples from the annotated samples to expand the dataset. Furthermore, we also create three automatically generated training and testing datasets of varying difficulty and distributions. Our extensive experiments show our proposed TRIGO poses a new challenge for advanced generative LM{'}s including GPT-4 which is pre-trained on a considerable amount of open-source formal theorem-proving language data, and provide a new tool to study the generative LM{'}s ability on both formal and mathematical reasoning.", }

Automated theorem proving (ATP) has become an appealing domain for exploring the reasoning ability of the recent successful generative language models. However, current ATP benchmarks are mainly focus on symbolic inference, but rarely involve the understanding of complex number combination reasoning. In this work, we propose TRIGO, an ATP benchmark that not only requires a model to reduce a trigonometric expression with step-by-step proof but also evaluates a generative LM{'}s reasoning ability on formulas and capability to manipulate, group, and factor number terms. We gather trigonometric expressions and their reduced forms from web, annotate the simplification process manually, and translate it into the {``}Lean{''} formal language system. We then automatically generate additional examples from the annotated samples to expand the dataset. Furthermore, we also create three automatically generated training and testing datasets of varying difficulty and distributions. Our extensive experiments show our proposed TRIGO poses a new challenge for advanced generative LM{'}s including GPT-4 which is pre-trained on a considerable amount of open-source formal theorem-proving language data, and provide a new tool to study the generative LM{'}s ability on both formal and mathematical reasoning.

TRIGO: Benchmarking Formal Mathematical Proof Reduction for Generative Language Models

emnlp-main.711

2310.10180

https://huggingface.co/papers/2310.10180

Oral

https://aclanthology.org/2023.emnlp-main.712.bib

https://aclanthology.org/2023.emnlp-main.712/

@inproceedings{mehandru-etal-2023-physician, title = "Physician Detection of Clinical Harm in Machine Translation: Quality Estimation Aids in Reliance and Backtranslation Identifies Critical Errors", author = "Mehandru, Nikita and Agrawal, Sweta and Xiao, Yimin and Gao, Ge and Khoong, Elaine and Carpuat, Marine and Salehi, Niloufar", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.712", doi = "10.18653/v1/2023.emnlp-main.712", pages = "11633--11647", abstract = "A major challenge in the practical use of Machine Translation (MT) is that users lack information on translation quality to make informed decisions about how to rely on outputs. Progress in quality estimation research provides techniques to automatically assess MT quality, but these techniques have primarily been evaluated in vitro by comparison against human judgments outside of a specific context of use. This paper evaluates quality estimation feedback in vivo with a human study in realistic high-stakes medical settings. Using Emergency Department discharge instructions, we study how interventions based on quality estimation versus backtranslation assist physicians in deciding whether to show MT outputs to a patient. We find that quality estimation improves appropriate reliance on MT, but backtranslation helps physicians detect more clinically harmful errors that QE alone often misses.", }

A major challenge in the practical use of Machine Translation (MT) is that users lack information on translation quality to make informed decisions about how to rely on outputs. Progress in quality estimation research provides techniques to automatically assess MT quality, but these techniques have primarily been evaluated in vitro by comparison against human judgments outside of a specific context of use. This paper evaluates quality estimation feedback in vivo with a human study in realistic high-stakes medical settings. Using Emergency Department discharge instructions, we study how interventions based on quality estimation versus backtranslation assist physicians in deciding whether to show MT outputs to a patient. We find that quality estimation improves appropriate reliance on MT, but backtranslation helps physicians detect more clinically harmful errors that QE alone often misses.

Physician Detection of Clinical Harm in Machine Translation: Quality Estimation Aids in Reliance and Backtranslation Identifies Critical Errors

emnlp-main.712

2310.16924

Poster

https://aclanthology.org/2023.emnlp-main.713.bib

https://aclanthology.org/2023.emnlp-main.713/

@inproceedings{weerasooriya-etal-2023-vicarious, title = "Vicarious Offense and Noise Audit of Offensive Speech Classifiers: Unifying Human and Machine Disagreement on What is Offensive", author = "Weerasooriya, Tharindu and Dutta, Sujan and Ranasinghe, Tharindu and Zampieri, Marcos and Homan, Christopher and KhudaBukhsh, Ashiqur", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.713", doi = "10.18653/v1/2023.emnlp-main.713", pages = "11648--11668", abstract = "Offensive speech detection is a key component of content moderation. However, what is offensive can be highly subjective. This paper investigates how machine and human moderators disagree on what is offensive when it comes to real-world social web political discourse. We show that (1) there is extensive disagreement among the moderators (humans and machines); and (2) human and large-language-model classifiers are unable to predict how other human raters will respond, based on their political leanings. For (1), we conduct a ***noise audit*** at an unprecedented scale that combines both machine and human responses. For (2), we introduce a first-of-its-kind dataset of ***vicarious offense***. Our noise audit reveals that moderation outcomes vary wildly across different machine moderators. Our experiments with human moderators suggest that political leanings combined with sensitive issues affect both first-person and vicarious offense. The dataset is available through https://github.com/Homan-Lab/voiced.", }

Offensive speech detection is a key component of content moderation. However, what is offensive can be highly subjective. This paper investigates how machine and human moderators disagree on what is offensive when it comes to real-world social web political discourse. We show that (1) there is extensive disagreement among the moderators (humans and machines); and (2) human and large-language-model classifiers are unable to predict how other human raters will respond, based on their political leanings. For (1), we conduct a ***noise audit*** at an unprecedented scale that combines both machine and human responses. For (2), we introduce a first-of-its-kind dataset of ***vicarious offense***. Our noise audit reveals that moderation outcomes vary wildly across different machine moderators. Our experiments with human moderators suggest that political leanings combined with sensitive issues affect both first-person and vicarious offense. The dataset is available through https://github.com/Homan-Lab/voiced.

Vicarious Offense and Noise Audit of Offensive Speech Classifiers: Unifying Human and Machine Disagreement on What is Offensive

emnlp-main.713

2301.12534

https://huggingface.co/papers/2301.12534

Poster

https://aclanthology.org/2023.emnlp-main.714.bib

https://aclanthology.org/2023.emnlp-main.714/

@inproceedings{ribeiro-etal-2023-generating, title = "Generating Summaries with Controllable Readability Levels", author = "Ribeiro, Leonardo F. R. and Bansal, Mohit and Dreyer, Markus", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.714", doi = "10.18653/v1/2023.emnlp-main.714", pages = "11669--11687", abstract = "Readability refers to how easily a reader can understand a written text. Several factors affect the readability level, such as the complexity of the text, its subject matter, and the reader{'}s background knowledge. Generating summaries based on different readability levels is critical for enabling knowledge consumption by diverse audiences. However, current text generation approaches lack refined control, resulting in texts that are not customized to readers{'} proficiency levels. In this work, we bridge this gap and study techniques to generate summaries at specified readability levels. Unlike previous methods that focus on a specific readability level (e.g., lay summarization), we generate summaries with fine-grained control over their readability. We develop three text generation techniques for controlling readability: (1) instruction-based readability control, (2) reinforcement learning to minimize the gap between requested and observed readability and (3) a decoding approach that uses lookahead to estimate the readability of upcoming decoding steps. We show that our generation methods significantly improve readability control on news summarization (CNN/DM dataset), as measured by various readability metrics and human judgement, establishing strong baselines for controllable readability in summarization.", }

Readability refers to how easily a reader can understand a written text. Several factors affect the readability level, such as the complexity of the text, its subject matter, and the reader{'}s background knowledge. Generating summaries based on different readability levels is critical for enabling knowledge consumption by diverse audiences. However, current text generation approaches lack refined control, resulting in texts that are not customized to readers{'} proficiency levels. In this work, we bridge this gap and study techniques to generate summaries at specified readability levels. Unlike previous methods that focus on a specific readability level (e.g., lay summarization), we generate summaries with fine-grained control over their readability. We develop three text generation techniques for controlling readability: (1) instruction-based readability control, (2) reinforcement learning to minimize the gap between requested and observed readability and (3) a decoding approach that uses lookahead to estimate the readability of upcoming decoding steps. We show that our generation methods significantly improve readability control on news summarization (CNN/DM dataset), as measured by various readability metrics and human judgement, establishing strong baselines for controllable readability in summarization.

Generating Summaries with Controllable Readability Levels

emnlp-main.714

2310.10623

https://huggingface.co/papers/2310.10623

Poster

https://aclanthology.org/2023.emnlp-main.715.bib

https://aclanthology.org/2023.emnlp-main.715/

@inproceedings{lin-etal-2023-maggretriever, title = "m{A}ggretriever: A Simple yet Effective Approach to Zero-Shot Multilingual Dense Retrieval", author = "Lin, Sheng-Chieh and Ahmad, Amin and Lin, Jimmy", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.715", doi = "10.18653/v1/2023.emnlp-main.715", pages = "11688--11696", abstract = "Multilingual information retrieval (MLIR) is a crucial yet challenging task due to the need for human annotations in multiple languages, making training data creation labor-intensive. In this paper, we introduce mAggretriever, which effectively leverages semantic and lexical features from pre-trained multilingual transformers (e.g., mBERT and XLM-R) for dense retrieval. To enhance training and inference efficiency, we employ approximate masked-language modeling prediction for computing lexical features, reducing 70{--}85{\%} GPU memory requirement for mAggretriever fine-tuning. Empirical results demonstrate that mAggretriever, fine-tuned solely on English training data, surpasses existing state-of-the-art multilingual dense retrieval models that undergo further training on large-scale MLIR training data. Our code is available at url.", }

Multilingual information retrieval (MLIR) is a crucial yet challenging task due to the need for human annotations in multiple languages, making training data creation labor-intensive. In this paper, we introduce mAggretriever, which effectively leverages semantic and lexical features from pre-trained multilingual transformers (e.g., mBERT and XLM-R) for dense retrieval. To enhance training and inference efficiency, we employ approximate masked-language modeling prediction for computing lexical features, reducing 70{--}85{\%} GPU memory requirement for mAggretriever fine-tuning. Empirical results demonstrate that mAggretriever, fine-tuned solely on English training data, surpasses existing state-of-the-art multilingual dense retrieval models that undergo further training on large-scale MLIR training data. Our code is available at url.

mAggretriever: A Simple yet Effective Approach to Zero-Shot Multilingual Dense Retrieval

emnlp-main.715

Poster

https://aclanthology.org/2023.emnlp-main.716.bib

https://aclanthology.org/2023.emnlp-main.716/

@inproceedings{singh-etal-2023-codefusion, title = "{C}ode{F}usion: A Pre-trained Diffusion Model for Code Generation", author = "Singh, Mukul and Cambronero, Jos{\'e} and Gulwani, Sumit and Le, Vu and Negreanu, Carina and Verbruggen, Gust", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.716", doi = "10.18653/v1/2023.emnlp-main.716", pages = "11697--11708", abstract = "Imagine a developer who can only change their last line of code{---}how often would they have to start writing a function from scratch before it is correct? Auto-regressive models for code generation from natural language have a similar limitation: they do not easily allow reconsidering earlier tokens generated. We introduce CodeFusion, a pre-trained diffusion code generation model that addresses this limitation by iteratively denoising a complete program conditioned on the encoded natural language. We evaluate CodeFusion on the task of natural language to code generation for Bash, Python, and Microsoft Excel conditional formatting (CF) rules. Experiments show that CodeFusion (75M parameters) performs on par with state-of-the-art auto-regressive systems (350M-175B parameters) in top-1 accuracy and outperforms them in top-3 and top-5 accuracy due to its better balance in diversity versus quality.", }

Imagine a developer who can only change their last line of code{---}how often would they have to start writing a function from scratch before it is correct? Auto-regressive models for code generation from natural language have a similar limitation: they do not easily allow reconsidering earlier tokens generated. We introduce CodeFusion, a pre-trained diffusion code generation model that addresses this limitation by iteratively denoising a complete program conditioned on the encoded natural language. We evaluate CodeFusion on the task of natural language to code generation for Bash, Python, and Microsoft Excel conditional formatting (CF) rules. Experiments show that CodeFusion (75M parameters) performs on par with state-of-the-art auto-regressive systems (350M-175B parameters) in top-1 accuracy and outperforms them in top-3 and top-5 accuracy due to its better balance in diversity versus quality.

CodeFusion: A Pre-trained Diffusion Model for Code Generation

emnlp-main.716

2310.17680

https://huggingface.co/papers/2310.17680

Oral

https://aclanthology.org/2023.emnlp-main.717.bib

https://aclanthology.org/2023.emnlp-main.717/

@inproceedings{aksu-etal-2023-cesar, title = "{CESAR}: Automatic Induction of Compositional Instructions for Multi-turn Dialogs", author = "Aksu, Taha and Hazarika, Devamanyu and Mehri, Shikib and Kim, Seokhwan and Hakkani-Tur, Dilek and Liu, Yang and Namazifar, Mahdi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.717", doi = "10.18653/v1/2023.emnlp-main.717", pages = "11709--11737", abstract = "Instruction-based multitasking has played a critical role in the success of large language models (LLMs) in multi-turn dialog applications. While publicly available LLMs have shown promising performance, when exposed to complex instructions with multiple constraints, they lag against state-of-the-art models like ChatGPT. In this work, we hypothesize that the availability of large-scale complex demonstrations is crucial in bridging this gap. Focusing on dialog applications, we propose a novel framework, CESAR, that unifies a large number of dialog tasks in the same format and allows programmatic induction of complex instructions without any manual effort. We apply CESAR on InstructDial, a benchmark for instruction-based dialog tasks. We further enhance InstructDial with new datasets and tasks and utilize CESAR to induce complex tasks with compositional instructions. This results in a new benchmark called InstructDial++, which includes 63 datasets with 86 basic tasks and 68 composite tasks. Through rigorous experiments, we demonstrate the scalability of CESAR in providing rich instructions. Models trained on InstructDial++ can follow compositional prompts, such as prompts that ask for multiple stylistic constraints.", }

Instruction-based multitasking has played a critical role in the success of large language models (LLMs) in multi-turn dialog applications. While publicly available LLMs have shown promising performance, when exposed to complex instructions with multiple constraints, they lag against state-of-the-art models like ChatGPT. In this work, we hypothesize that the availability of large-scale complex demonstrations is crucial in bridging this gap. Focusing on dialog applications, we propose a novel framework, CESAR, that unifies a large number of dialog tasks in the same format and allows programmatic induction of complex instructions without any manual effort. We apply CESAR on InstructDial, a benchmark for instruction-based dialog tasks. We further enhance InstructDial with new datasets and tasks and utilize CESAR to induce complex tasks with compositional instructions. This results in a new benchmark called InstructDial++, which includes 63 datasets with 86 basic tasks and 68 composite tasks. Through rigorous experiments, we demonstrate the scalability of CESAR in providing rich instructions. Models trained on InstructDial++ can follow compositional prompts, such as prompts that ask for multiple stylistic constraints.

CESAR: Automatic Induction of Compositional Instructions for Multi-turn Dialogs

emnlp-main.717

2311.17376

Poster

https://aclanthology.org/2023.emnlp-main.718.bib

https://aclanthology.org/2023.emnlp-main.718/

@inproceedings{xu-etal-2023-vechr, title = "{VECHR}: A Dataset for Explainable and Robust Classification of Vulnerability Type in the {E}uropean Court of Human Rights", author = "Xu, Shanshan and Staufer, Leon and T.y.s.s, Santosh and Ichim, Oana and Heri, Corina and Grabmair, Matthias", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.718", doi = "10.18653/v1/2023.emnlp-main.718", pages = "11738--11752", abstract = "Recognizing vulnerability is crucial for understanding and implementing targeted support to empower individuals in need. This is especially important at the European Court of Human Rights (ECtHR), where the court adapts Convention standards to meet actual individual needs and thus to ensure effective human rights protection. However, the concept of vulnerability remains elusive at the ECtHR and no prior NLP research has dealt with it. To enable future research in this area, we present VECHR, a novel expert-annotated multi-label dataset comprising of vulnerability type classification and explanation rationale. We benchmark the performance of state-of-the-art models on VECHR from both prediction and explainability perspective. Our results demonstrate the challenging nature of task with lower prediction performance and limited agreement between models and experts. Further, we analyze the robustness of these models in dealing with out-of-domain (OOD) data and observe overall limited performance. Our dataset poses unique challenges offering a significant room for improvement regarding performance, explainability and robustness.", }

Recognizing vulnerability is crucial for understanding and implementing targeted support to empower individuals in need. This is especially important at the European Court of Human Rights (ECtHR), where the court adapts Convention standards to meet actual individual needs and thus to ensure effective human rights protection. However, the concept of vulnerability remains elusive at the ECtHR and no prior NLP research has dealt with it. To enable future research in this area, we present VECHR, a novel expert-annotated multi-label dataset comprising of vulnerability type classification and explanation rationale. We benchmark the performance of state-of-the-art models on VECHR from both prediction and explainability perspective. Our results demonstrate the challenging nature of task with lower prediction performance and limited agreement between models and experts. Further, we analyze the robustness of these models in dealing with out-of-domain (OOD) data and observe overall limited performance. Our dataset poses unique challenges offering a significant room for improvement regarding performance, explainability and robustness.

VECHR: A Dataset for Explainable and Robust Classification of Vulnerability Type in the European Court of Human Rights

emnlp-main.718

2310.11368

https://huggingface.co/papers/2310.11368

Poster

https://aclanthology.org/2023.emnlp-main.719.bib

https://aclanthology.org/2023.emnlp-main.719/

@inproceedings{wu-etal-2023-acquired, title = "{ACQUIRED}: A Dataset for Answering Counterfactual Questions In Real-Life Videos", author = "Wu, Te-Lin and Dou, Zi-Yi and Hu, Qingyuan and Hou, Yu and Chandra, Nischal and Freedman, Marjorie and Weischedel, Ralph and Peng, Nanyun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.719", doi = "10.18653/v1/2023.emnlp-main.719", pages = "11753--11770", abstract = "Multimodal counterfactual reasoning is a vital yet challenging ability for AI systems. It involves predicting the outcomes of hypothetical circumstances based on vision and language inputs, which enables AI models to learn from failures and explore hypothetical scenarios. Despite its importance, there are only a few datasets targeting the counterfactual reasoning abilities of multimodal models. Among them, they only cover reasoning over synthetic environments or specific types of events (e.g. traffic collisions), making them hard to reliably benchmark the model generalization ability in diverse real-world scenarios and reasoning dimensions. To overcome these limitations, we develop a video question answering dataset, ACQUIRED: it consists of 3.9K annotated videos, encompassing a wide range of event types and incorporating both first and third-person viewpoints, which ensures a focus on real-world diversity. In addition, each video is annotated with questions that span three distinct dimensions of reasoning, including physical, social, and temporal, which can comprehensively evaluate the model counterfactual abilities along multiple aspects. We benchmark our dataset against several state-of-the-art language-only and multimodal models and experimental results demonstrate a significant performance gap ({\textgreater}13{\%}) between models and humans. The findings suggest that multimodal counterfactual reasoning remains an open challenge and ACQUIRED is a comprehensive and reliable benchmark for inspiring future research in this direction.", }

Multimodal counterfactual reasoning is a vital yet challenging ability for AI systems. It involves predicting the outcomes of hypothetical circumstances based on vision and language inputs, which enables AI models to learn from failures and explore hypothetical scenarios. Despite its importance, there are only a few datasets targeting the counterfactual reasoning abilities of multimodal models. Among them, they only cover reasoning over synthetic environments or specific types of events (e.g. traffic collisions), making them hard to reliably benchmark the model generalization ability in diverse real-world scenarios and reasoning dimensions. To overcome these limitations, we develop a video question answering dataset, ACQUIRED: it consists of 3.9K annotated videos, encompassing a wide range of event types and incorporating both first and third-person viewpoints, which ensures a focus on real-world diversity. In addition, each video is annotated with questions that span three distinct dimensions of reasoning, including physical, social, and temporal, which can comprehensively evaluate the model counterfactual abilities along multiple aspects. We benchmark our dataset against several state-of-the-art language-only and multimodal models and experimental results demonstrate a significant performance gap ({\textgreater}13{\%}) between models and humans. The findings suggest that multimodal counterfactual reasoning remains an open challenge and ACQUIRED is a comprehensive and reliable benchmark for inspiring future research in this direction.

ACQUIRED: A Dataset for Answering Counterfactual Questions In Real-Life Videos

emnlp-main.719

2311.01620

https://huggingface.co/papers/2311.01620

Poster

https://aclanthology.org/2023.emnlp-main.720.bib

https://aclanthology.org/2023.emnlp-main.720/

@inproceedings{guo-etal-2023-parse, title = "From Parse-Execute to Parse-Execute-Refine: Improving Semantic Parser for Complex Question Answering over Knowledge Base", author = "Guo, Wangzhen and Luo, Linyin and Lai, Hanjiang and Yin, Jian", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.720", doi = "10.18653/v1/2023.emnlp-main.720", pages = "11771--11780", abstract = "Parsing questions into executable logical forms has showed impressive results for knowledge-base question answering (KBQA). However, complex KBQA is a more challenging task that requires to perform complex multi-step reasoning. Recently, a new semantic parser called KoPL has been proposed to explicitly model the reasoning processes, which achieved the state-of-the-art on complex KBQA. In this paper, we further explore how to unlock the reasoning ability of semantic parsers by a simple proposed parse-execute-refine paradigm. We refine and improve the KoPL parser by demonstrating the executed intermediate reasoning steps to the KBQA model. We show that such simple strategy can significantly improve the ability of complex reasoning. Specifically, we propose three components: a parsing stage, an execution stage and a refinement stage, to enhance the ability of complex reasoning. The parser uses the KoPL to generate the transparent logical forms. Then, the execution stage aligns and executes the logical forms over knowledge base to obtain intermediate reasoning processes. Finally, the intermediate step-by-step reasoning processes are demonstrated to the KBQA model in the refinement stage. With the explicit reasoning processes, it is much easier to answer the complex questions. Experiments on benchmark dataset shows that the proposed PER-KBQA performs significantly better than the stage-of-the-art baselines on the complex KBQA.", }

Parsing questions into executable logical forms has showed impressive results for knowledge-base question answering (KBQA). However, complex KBQA is a more challenging task that requires to perform complex multi-step reasoning. Recently, a new semantic parser called KoPL has been proposed to explicitly model the reasoning processes, which achieved the state-of-the-art on complex KBQA. In this paper, we further explore how to unlock the reasoning ability of semantic parsers by a simple proposed parse-execute-refine paradigm. We refine and improve the KoPL parser by demonstrating the executed intermediate reasoning steps to the KBQA model. We show that such simple strategy can significantly improve the ability of complex reasoning. Specifically, we propose three components: a parsing stage, an execution stage and a refinement stage, to enhance the ability of complex reasoning. The parser uses the KoPL to generate the transparent logical forms. Then, the execution stage aligns and executes the logical forms over knowledge base to obtain intermediate reasoning processes. Finally, the intermediate step-by-step reasoning processes are demonstrated to the KBQA model in the refinement stage. With the explicit reasoning processes, it is much easier to answer the complex questions. Experiments on benchmark dataset shows that the proposed PER-KBQA performs significantly better than the stage-of-the-art baselines on the complex KBQA.

From Parse-Execute to Parse-Execute-Refine: Improving Semantic Parser for Complex Question Answering over Knowledge Base

emnlp-main.720

2305.03356

Poster

https://aclanthology.org/2023.emnlp-main.721.bib

https://aclanthology.org/2023.emnlp-main.721/

@inproceedings{deng-raffel-2023-reward, title = "Reward-Augmented Decoding: Efficient Controlled Text Generation With a Unidirectional Reward Model", author = "Deng, Haikang and Raffel, Colin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.721", doi = "10.18653/v1/2023.emnlp-main.721", pages = "11781--11791", abstract = "While large language models have proven effective in a huge range of downstream applications, they often generate text that is problematic or lacks a desired attribute. In this paper, we introduce Reward-Augmented Decoding (RAD), a text generation procedure that uses a small unidirectional reward model to encourage a language model to generate text that has certain properties. Specifically, RAD uses the reward model to score generations as they are produced and rescales sampling probabilities to favor high-reward tokens. By using a unidirectional reward model, RAD can cache activations from prior generation steps to decrease computational overhead. Through experiments on generating non-toxic and sentiment-controlled text, we demonstrate that RAD performs best among methods that change only the generation procedure and matches the performance of state-of-the-art methods that involve re-training the language model. We further validate that RAD is effective on very large language models while incurring a minimal computational overhead.", }

While large language models have proven effective in a huge range of downstream applications, they often generate text that is problematic or lacks a desired attribute. In this paper, we introduce Reward-Augmented Decoding (RAD), a text generation procedure that uses a small unidirectional reward model to encourage a language model to generate text that has certain properties. Specifically, RAD uses the reward model to score generations as they are produced and rescales sampling probabilities to favor high-reward tokens. By using a unidirectional reward model, RAD can cache activations from prior generation steps to decrease computational overhead. Through experiments on generating non-toxic and sentiment-controlled text, we demonstrate that RAD performs best among methods that change only the generation procedure and matches the performance of state-of-the-art methods that involve re-training the language model. We further validate that RAD is effective on very large language models while incurring a minimal computational overhead.

Reward-Augmented Decoding: Efficient Controlled Text Generation With a Unidirectional Reward Model

emnlp-main.721

2310.09520

https://huggingface.co/papers/2310.09520

Poster

https://aclanthology.org/2023.emnlp-main.722.bib

https://aclanthology.org/2023.emnlp-main.722/

@inproceedings{borchert-etal-2023-core, title = "{CORE}: A Few-Shot Company Relation Classification Dataset for Robust Domain Adaptation.", author = "Borchert, Philipp and De Weerdt, Jochen and Coussement, Kristof and De Caigny, Arno and Moens, Marie-Francine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.722", doi = "10.18653/v1/2023.emnlp-main.722", pages = "11792--11806", abstract = "We introduce CORE, a dataset for few-shot relation classification (RC) focused on company relations and business entities. CORE includes 4,708 instances of 12 relation types with corresponding textual evidence extracted from company Wikipedia pages. Company names and business entities pose a challenge for few-shot RC models due to the rich and diverse information associated with them. For example, a company name may represent the legal entity, products, people, or business divisions depending on the context. Therefore, deriving the relation type between entities is highly dependent on textual context. To evaluate the performance of state-of-the-art RC models on the CORE dataset, we conduct experiments in the few-shot domain adaptation setting. Our results reveal substantial performance gaps, confirming that models trained on different domains struggle to adapt to CORE. Interestingly, we find that models trained on CORE showcase improved out-of-domain performance, which highlights the importance of high-quality data for robust domain generalization. Specifically, the information richness embedded in business entities allows models to focus on contextual nuances, reducing their reliance on superficial clues such as relation-specific verbs. In addition to the dataset, we provide relevant code snippets to facilitate reproducibility and encourage further research in the field. The CORE dataset and code are publicly available at \url{https://anonymous.4open.science/r/CORE-D377}.", }

We introduce CORE, a dataset for few-shot relation classification (RC) focused on company relations and business entities. CORE includes 4,708 instances of 12 relation types with corresponding textual evidence extracted from company Wikipedia pages. Company names and business entities pose a challenge for few-shot RC models due to the rich and diverse information associated with them. For example, a company name may represent the legal entity, products, people, or business divisions depending on the context. Therefore, deriving the relation type between entities is highly dependent on textual context. To evaluate the performance of state-of-the-art RC models on the CORE dataset, we conduct experiments in the few-shot domain adaptation setting. Our results reveal substantial performance gaps, confirming that models trained on different domains struggle to adapt to CORE. Interestingly, we find that models trained on CORE showcase improved out-of-domain performance, which highlights the importance of high-quality data for robust domain generalization. Specifically, the information richness embedded in business entities allows models to focus on contextual nuances, reducing their reliance on superficial clues such as relation-specific verbs. In addition to the dataset, we provide relevant code snippets to facilitate reproducibility and encourage further research in the field. The CORE dataset and code are publicly available at \url{https://anonymous.4open.science/r/CORE-D377}.

CORE: A Few-Shot Company Relation Classification Dataset for Robust Domain Adaptation.

emnlp-main.722

2310.12024

https://huggingface.co/papers/2310.12024

Oral

https://aclanthology.org/2023.emnlp-main.723.bib

https://aclanthology.org/2023.emnlp-main.723/

@inproceedings{liu-wan-2023-models, title = "Models See Hallucinations: Evaluating the Factuality in Video Captioning", author = "Liu, Hui and Wan, Xiaojun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.723", doi = "10.18653/v1/2023.emnlp-main.723", pages = "11807--11823", abstract = "Video captioning aims to describe events in a video with natural language. In recent years, many works have focused on improving captioning models{'} performance. However, like other text generation tasks, it risks introducing factual errors not supported by the input video. Factual errors can seriously affect the quality of the generated text, sometimes making it completely unusable. Although factual consistency has received much research attention in text-to-text tasks (e.g., summarization), it is less studied in vision-based text generation. In this work, we conduct the first human evaluation of the factuality in video captioning and annotate two factuality datasets. We find that 56{\%} of the model-generated sentences have factual errors, indicating it is a severe problem in this field, but existing evaluation metrics show little correlation with human factuality annotation. We further propose a weakly-supervised, model-based factuality metric FactVC, which outperforms previous metrics on factuality evaluation of video captioning.", }

Video captioning aims to describe events in a video with natural language. In recent years, many works have focused on improving captioning models{'} performance. However, like other text generation tasks, it risks introducing factual errors not supported by the input video. Factual errors can seriously affect the quality of the generated text, sometimes making it completely unusable. Although factual consistency has received much research attention in text-to-text tasks (e.g., summarization), it is less studied in vision-based text generation. In this work, we conduct the first human evaluation of the factuality in video captioning and annotate two factuality datasets. We find that 56{\%} of the model-generated sentences have factual errors, indicating it is a severe problem in this field, but existing evaluation metrics show little correlation with human factuality annotation. We further propose a weakly-supervised, model-based factuality metric FactVC, which outperforms previous metrics on factuality evaluation of video captioning.

Models See Hallucinations: Evaluating the Factuality in Video Captioning

emnlp-main.723

2303.02961

Oral

https://aclanthology.org/2023.emnlp-main.724.bib

https://aclanthology.org/2023.emnlp-main.724/

@inproceedings{kubis-etal-2023-back, title = "Back Transcription as a Method for Evaluating Robustness of Natural Language Understanding Models to Speech Recognition Errors", author = "Kubis, Marek and Sk{\'o}rzewski, Pawe{\l} and Sowa{\'n}ski, Marcin and Zietkiewicz, Tomasz", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.724", doi = "10.18653/v1/2023.emnlp-main.724", pages = "11824--11835", abstract = "In a spoken dialogue system, an NLU model is preceded by a speech recognition system that can deteriorate the performance of natural language understanding. This paper proposes a method for investigating the impact of speech recognition errors on the performance of natural language understanding models. The proposed method combines the back transcription procedure with a fine-grained technique for categorizing the errors that affect the performance of NLU models. The method relies on the usage of synthesized speech for NLU evaluation. We show that the use of synthesized speech in place of audio recording does not change the outcomes of the presented technique in a significant way.", }

In a spoken dialogue system, an NLU model is preceded by a speech recognition system that can deteriorate the performance of natural language understanding. This paper proposes a method for investigating the impact of speech recognition errors on the performance of natural language understanding models. The proposed method combines the back transcription procedure with a fine-grained technique for categorizing the errors that affect the performance of NLU models. The method relies on the usage of synthesized speech for NLU evaluation. We show that the use of synthesized speech in place of audio recording does not change the outcomes of the presented technique in a significant way.

Back Transcription as a Method for Evaluating Robustness of Natural Language Understanding Models to Speech Recognition Errors

emnlp-main.724

2310.16609

https://huggingface.co/papers/2310.16609

Poster

https://aclanthology.org/2023.emnlp-main.725.bib

https://aclanthology.org/2023.emnlp-main.725/

@inproceedings{chatterjee-etal-2023-cabbage, title = "Cabbage Sweeter than Cake? Analysing the Potential of Large Language Models for Learning Conceptual Spaces", author = "Chatterjee, Usashi and Gajbhiye, Amit and Schockaert, Steven", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.725", doi = "10.18653/v1/2023.emnlp-main.725", pages = "11836--11842", abstract = "The theory of Conceptual Spaces is an influential cognitive-linguistic framework for representing the meaning of concepts. Conceptual spaces are constructed from a set of quality dimensions, which essentially correspond to primitive perceptual features (e.g. hue or size). These quality dimensions are usually learned from human judgements, which means that applications of conceptual spaces tend to be limited to narrow domains (e.g. modelling colour or taste). Encouraged by recent findings about the ability of Large Language Models (LLMs) to learn perceptually grounded representations, we explore the potential of such models for learning conceptual spaces. Our experiments show that LLMs can indeed be used for learning meaningful representations to some extent. However, we also find that fine-tuned models of the BERT family are able to match or even outperform the largest GPT-3 model, despite being 2 to 3 orders of magnitude smaller.", }

The theory of Conceptual Spaces is an influential cognitive-linguistic framework for representing the meaning of concepts. Conceptual spaces are constructed from a set of quality dimensions, which essentially correspond to primitive perceptual features (e.g. hue or size). These quality dimensions are usually learned from human judgements, which means that applications of conceptual spaces tend to be limited to narrow domains (e.g. modelling colour or taste). Encouraged by recent findings about the ability of Large Language Models (LLMs) to learn perceptually grounded representations, we explore the potential of such models for learning conceptual spaces. Our experiments show that LLMs can indeed be used for learning meaningful representations to some extent. However, we also find that fine-tuned models of the BERT family are able to match or even outperform the largest GPT-3 model, despite being 2 to 3 orders of magnitude smaller.

Cabbage Sweeter than Cake? Analysing the Potential of Large Language Models for Learning Conceptual Spaces

emnlp-main.725

2310.05481

Poster

https://aclanthology.org/2023.emnlp-main.726.bib

https://aclanthology.org/2023.emnlp-main.726/

@inproceedings{li-etal-2023-language-models, title = "Can Language Models Understand Physical Concepts?", author = "Li, Lei and Xu, Jingjing and Dong, Qingxiu and Zheng, Ce and Sun, Xu and Kong, Lingpeng and Liu, Qi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.726", doi = "10.18653/v1/2023.emnlp-main.726", pages = "11843--11861", abstract = "Language models (LMs) gradually become general-purpose interfaces in the interactive and embodied world, where the understanding of physical concepts is an essential prerequisite. However, it is unclear whether LMs can understand physical concepts in the human world. To investigate this, we design a benchmark VEC that covers the tasks of (i) Visual concepts, such as the shape and material of objects, and (ii) Embodied Concepts, learned from the interaction with the world such as the temperature of objects. Our zero (few)-shot prompting results show that the understanding of certain visual concepts emerges as scaling up LMs, but there are still basic concepts to which the scaling law does not apply. For example, OPT-175B performs close to humans with a zero-shot accuracy of 85{\%} on the material concept, yet behaves like random guessing on the mass concept. Instead, vision-augmented LMs such as CLIP and BLIP achieve a human-level understanding of embodied concepts. Analysis indicates that the rich semantics in visual representation can serve as a valuable source of embodied knowledge. Inspired by this, we propose a distillation method to transfer embodied knowledge from VLMs to LMs, achieving performance gain comparable with that by scaling up parameters of LMs $134\times$. Our dataset is available at https://github.com/TobiasLee/VEC.", }

Language models (LMs) gradually become general-purpose interfaces in the interactive and embodied world, where the understanding of physical concepts is an essential prerequisite. However, it is unclear whether LMs can understand physical concepts in the human world. To investigate this, we design a benchmark VEC that covers the tasks of (i) Visual concepts, such as the shape and material of objects, and (ii) Embodied Concepts, learned from the interaction with the world such as the temperature of objects. Our zero (few)-shot prompting results show that the understanding of certain visual concepts emerges as scaling up LMs, but there are still basic concepts to which the scaling law does not apply. For example, OPT-175B performs close to humans with a zero-shot accuracy of 85{\%} on the material concept, yet behaves like random guessing on the mass concept. Instead, vision-augmented LMs such as CLIP and BLIP achieve a human-level understanding of embodied concepts. Analysis indicates that the rich semantics in visual representation can serve as a valuable source of embodied knowledge. Inspired by this, we propose a distillation method to transfer embodied knowledge from VLMs to LMs, achieving performance gain comparable with that by scaling up parameters of LMs $134\times$. Our dataset is available at https://github.com/TobiasLee/VEC.

Can Language Models Understand Physical Concepts?

emnlp-main.726

Poster

https://aclanthology.org/2023.emnlp-main.727.bib

https://aclanthology.org/2023.emnlp-main.727/

@inproceedings{zhu-tan-2023-spt, title = "{SPT}: Learning to Selectively Insert Prompts for Better Prompt Tuning", author = "Zhu, Wei and Tan, Ming", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.727", doi = "10.18653/v1/2023.emnlp-main.727", pages = "11862--11878", abstract = "Prompt tuning prepends a soft prompt to the input embeddings or hidden states and only optimizes the prompt to adapt pretrained models (PTMs) to downstream tasks. The previous work manually selects prompt layers which are far from optimal and failed to exploit the potential of prompt tuning. In this work, we propose a novel framework, Selective Prompt Tuning (SPT), that learns to select the proper prompt layers by inserting a prompt controlled by a learnable probabilistic gate at each intermediate layer. We further propose a novel bi-level optimization framework, SPT-DARTS, that can better optimize the learnable gates and improve the final prompt tuning performances of the learned prompt layer settings. We conduct extensive experiments with ten benchmark datasets under the full-data and few-shot scenarios. The results demonstrate that our SPT framework can perform better than the previous state-of-the-art PETuning baselines with comparable or fewer tunable parameters.", }

Prompt tuning prepends a soft prompt to the input embeddings or hidden states and only optimizes the prompt to adapt pretrained models (PTMs) to downstream tasks. The previous work manually selects prompt layers which are far from optimal and failed to exploit the potential of prompt tuning. In this work, we propose a novel framework, Selective Prompt Tuning (SPT), that learns to select the proper prompt layers by inserting a prompt controlled by a learnable probabilistic gate at each intermediate layer. We further propose a novel bi-level optimization framework, SPT-DARTS, that can better optimize the learnable gates and improve the final prompt tuning performances of the learned prompt layer settings. We conduct extensive experiments with ten benchmark datasets under the full-data and few-shot scenarios. The results demonstrate that our SPT framework can perform better than the previous state-of-the-art PETuning baselines with comparable or fewer tunable parameters.

SPT: Learning to Selectively Insert Prompts for Better Prompt Tuning

emnlp-main.727

Poster

https://aclanthology.org/2023.emnlp-main.728.bib

https://aclanthology.org/2023.emnlp-main.728/

@inproceedings{yang-etal-2023-upon, title = "Once Upon a ${\it Time}$ in ${\it Graph}$: Relative-Time Pretraining for Complex Temporal Reasoning", author = "Yang, Sen and Li, Xin and Bing, Lidong and Lam, Wai", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.728", doi = "10.18653/v1/2023.emnlp-main.728", pages = "11879--11895", abstract = "Our physical world is constantly evolving over time, rendering challenges for pre-trained language models to understand and reason over the temporal contexts of texts. Existing work focuses on strengthening the direct association between a piece of text and its time-stamp. However, the knowledge-time association is usually insufficient for the downstream tasks that require reasoning over temporal dependencies between knowledge. In this work, we make use of the underlying nature of time, all temporally-scoped sentences are strung together through a one-dimensional time axis, and suggest creating a graph structure based on the relative placements of events along the time axis. Inspired by the graph view, we propose RemeMo ($\underline{Re}lative Ti\underline{me} \underline{Mo}deling$), which explicitly connects all temporally-scoped facts by modeling the time relations between any two sentences. Experimental results show that RemeMo outperforms the baseline T5 on multiple temporal question answering datasets under various settings. Further analysis suggests that RemeMo is especially good at modeling long-range complex temporal dependencies.", }

Our physical world is constantly evolving over time, rendering challenges for pre-trained language models to understand and reason over the temporal contexts of texts. Existing work focuses on strengthening the direct association between a piece of text and its time-stamp. However, the knowledge-time association is usually insufficient for the downstream tasks that require reasoning over temporal dependencies between knowledge. In this work, we make use of the underlying nature of time, all temporally-scoped sentences are strung together through a one-dimensional time axis, and suggest creating a graph structure based on the relative placements of events along the time axis. Inspired by the graph view, we propose RemeMo ($\underline{Re}lative Ti\underline{me} \underline{Mo}deling$), which explicitly connects all temporally-scoped facts by modeling the time relations between any two sentences. Experimental results show that RemeMo outperforms the baseline T5 on multiple temporal question answering datasets under various settings. Further analysis suggests that RemeMo is especially good at modeling long-range complex temporal dependencies.

Once Upon a Time in Graph: Relative-Time Pretraining for Complex Temporal Reasoning

emnlp-main.728

2310.14709

Oral

https://aclanthology.org/2023.emnlp-main.729.bib

https://aclanthology.org/2023.emnlp-main.729/

@inproceedings{balepur-etal-2023-expository, title = "Expository Text Generation: Imitate, Retrieve, Paraphrase", author = "Balepur, Nishant and Huang, Jie and Chang, Kevin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.729", doi = "10.18653/v1/2023.emnlp-main.729", pages = "11896--11919", abstract = "Expository documents are vital resources for conveying complex information to readers. Despite their usefulness, writing expository text by hand is a challenging process that requires careful content planning, obtaining facts from multiple sources, and the ability to clearly synthesize these facts. To ease these burdens, we propose the task of expository text generation, which seeks to automatically generate an accurate and stylistically consistent expository text for a topic by intelligently searching a knowledge source. We solve our task by developing IRP, a framework that overcomes the limitations of retrieval-augmented models and iteratively performs content planning, fact retrieval, and rephrasing. Through experiments on three diverse, newly-collected datasets, we show that IRP produces factual and organized expository texts that accurately inform readers.", }

Expository documents are vital resources for conveying complex information to readers. Despite their usefulness, writing expository text by hand is a challenging process that requires careful content planning, obtaining facts from multiple sources, and the ability to clearly synthesize these facts. To ease these burdens, we propose the task of expository text generation, which seeks to automatically generate an accurate and stylistically consistent expository text for a topic by intelligently searching a knowledge source. We solve our task by developing IRP, a framework that overcomes the limitations of retrieval-augmented models and iteratively performs content planning, fact retrieval, and rephrasing. Through experiments on three diverse, newly-collected datasets, we show that IRP produces factual and organized expository texts that accurately inform readers.

Expository Text Generation: Imitate, Retrieve, Paraphrase

emnlp-main.729

2305.03276

https://huggingface.co/papers/2305.03276

Poster

https://aclanthology.org/2023.emnlp-main.730.bib

https://aclanthology.org/2023.emnlp-main.730/

@inproceedings{laouar-etal-2023-large, title = "Large-scale similarity search with Optimal Transport", author = "Laouar, Cl{\'e}a and Takezawa, Yuki and Yamada, Makoto", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.730", doi = "10.18653/v1/2023.emnlp-main.730", pages = "11920--11930", abstract = "Wasserstein distance is a powerful tool for comparing probability distributions and is widely used for document classification and retrieval tasks in NLP. In particular, it is known as the word mover{'}s distance (WMD) in the NLP community. WMD exhibits excellent performance for various NLP tasks; however, one of its limitations is its computational cost and thus is not useful for large-scale distribution comparisons. In this study, we propose a simple and effective nearest neighbor search based on the Wasserstein distance. Specifically, we employ the L1 embedding method based on the tree-based Wasserstein approximation and subsequently used the nearest neighbor search to efficiently find the $k$-nearest neighbors. Through benchmark experiments, we demonstrate that the proposed approximation has comparable performance to the vanilla Wasserstein distance and can be computed three orders of magnitude faster than the vanilla Wasserstein distance.", }

Wasserstein distance is a powerful tool for comparing probability distributions and is widely used for document classification and retrieval tasks in NLP. In particular, it is known as the word mover{'}s distance (WMD) in the NLP community. WMD exhibits excellent performance for various NLP tasks; however, one of its limitations is its computational cost and thus is not useful for large-scale distribution comparisons. In this study, we propose a simple and effective nearest neighbor search based on the Wasserstein distance. Specifically, we employ the L1 embedding method based on the tree-based Wasserstein approximation and subsequently used the nearest neighbor search to efficiently find the $k$-nearest neighbors. Through benchmark experiments, we demonstrate that the proposed approximation has comparable performance to the vanilla Wasserstein distance and can be computed three orders of magnitude faster than the vanilla Wasserstein distance.

Large-scale similarity search with Optimal Transport

emnlp-main.730

Poster

https://aclanthology.org/2023.emnlp-main.731.bib

https://aclanthology.org/2023.emnlp-main.731/

@inproceedings{singh-etal-2023-enhancing, title = "Enhancing Textbooks with Visuals from the Web for Improved Learning", author = "Singh, Janvijay and Zouhar, Vil{\'e}m and Sachan, Mrinmaya", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.731", doi = "10.18653/v1/2023.emnlp-main.731", pages = "11931--11944", abstract = "Textbooks are one of the main mediums for delivering high-quality education to students. In particular, explanatory and illustrative visuals play a key role in retention, comprehension and general transfer of knowledge. However, many textbooks lack these interesting visuals to support student learning. In this paper, we investigate the effectiveness of vision-language models to automatically enhance textbooks with images from the web. We collect a dataset of e-textbooks in the math, science, social science and business domains. We then set up a text-image matching task that involves retrieving and appropriately assigning web images to textbooks, which we frame as a matching optimization problem. Through a crowd-sourced evaluation, we verify that (1) while the original textbook images are rated higher, automatically assigned ones are not far behind, and (2) the precise formulation of the optimization problem matters. We release the dataset of textbooks with an associated image bank to inspire further research in this intersectional area of computer vision and NLP for education.", }

Textbooks are one of the main mediums for delivering high-quality education to students. In particular, explanatory and illustrative visuals play a key role in retention, comprehension and general transfer of knowledge. However, many textbooks lack these interesting visuals to support student learning. In this paper, we investigate the effectiveness of vision-language models to automatically enhance textbooks with images from the web. We collect a dataset of e-textbooks in the math, science, social science and business domains. We then set up a text-image matching task that involves retrieving and appropriately assigning web images to textbooks, which we frame as a matching optimization problem. Through a crowd-sourced evaluation, we verify that (1) while the original textbook images are rated higher, automatically assigned ones are not far behind, and (2) the precise formulation of the optimization problem matters. We release the dataset of textbooks with an associated image bank to inspire further research in this intersectional area of computer vision and NLP for education.

Enhancing Textbooks with Visuals from the Web for Improved Learning

emnlp-main.731

2304.08931

Poster

https://aclanthology.org/2023.emnlp-main.732.bib

https://aclanthology.org/2023.emnlp-main.732/

@inproceedings{wang-etal-2023-continual, title = "Continual Event Extraction with Semantic Confusion Rectification", author = "Wang, Zitao and Wang, Xinyi and Hu, Wei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.732", doi = "10.18653/v1/2023.emnlp-main.732", pages = "11945--11955", abstract = "We study continual event extraction, which aims to extract incessantly emerging event information while avoiding forgetting. We observe that the semantic confusion on event types stems from the annotations of the same text being updated over time. The imbalance between event types even aggravates this issue. This paper proposes a novel continual event extraction model with semantic confusion rectification. We mark pseudo labels for each sentence to alleviate semantic confusion. We transfer pivotal knowledge between current and previous models to enhance the understanding of event types. Moreover, we encourage the model to focus on the semantics of long-tailed event types by leveraging other associated types. Experimental results show that our model outperforms state-of-the-art baselines and is proficient in imbalanced datasets.", }

We study continual event extraction, which aims to extract incessantly emerging event information while avoiding forgetting. We observe that the semantic confusion on event types stems from the annotations of the same text being updated over time. The imbalance between event types even aggravates this issue. This paper proposes a novel continual event extraction model with semantic confusion rectification. We mark pseudo labels for each sentence to alleviate semantic confusion. We transfer pivotal knowledge between current and previous models to enhance the understanding of event types. Moreover, we encourage the model to focus on the semantics of long-tailed event types by leveraging other associated types. Experimental results show that our model outperforms state-of-the-art baselines and is proficient in imbalanced datasets.

Continual Event Extraction with Semantic Confusion Rectification

emnlp-main.732

2310.15470

Oral

https://aclanthology.org/2023.emnlp-main.733.bib

https://aclanthology.org/2023.emnlp-main.733/

@inproceedings{farinhas-etal-2023-empirical, title = "An Empirical Study of Translation Hypothesis Ensembling with Large Language Models", author = "Farinhas, Ant{\'o}nio and de Souza, Jos{\'e} and Martins, Andre", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.733", doi = "10.18653/v1/2023.emnlp-main.733", pages = "11956--11970", abstract = "Large language models (LLMs) are becoming a one-fits-many solution, but they sometimes hallucinate or produce unreliable output. In this paper, we investigate how hypothesis ensembling can improve the quality of the generated text for the specific problem of LLM-based machine translation. We experiment with several techniques for ensembling hypotheses produced by LLMs such as ChatGPT, LLaMA, and Alpaca. We provide a comprehensive study along multiple dimensions, including the method to generate hypotheses (multiple prompts, temperature-based sampling, and beam search) and the strategy to produce the final translation (instruction-based, quality-based reranking, and minimum Bayes risk (MBR) decoding). Our results show that MBR decoding is a very effective method, that translation quality can be improved using a small number of samples, and that instruction tuning has a strong impact on the relation between the diversity of the hypotheses and the sampling temperature.", }

Large language models (LLMs) are becoming a one-fits-many solution, but they sometimes hallucinate or produce unreliable output. In this paper, we investigate how hypothesis ensembling can improve the quality of the generated text for the specific problem of LLM-based machine translation. We experiment with several techniques for ensembling hypotheses produced by LLMs such as ChatGPT, LLaMA, and Alpaca. We provide a comprehensive study along multiple dimensions, including the method to generate hypotheses (multiple prompts, temperature-based sampling, and beam search) and the strategy to produce the final translation (instruction-based, quality-based reranking, and minimum Bayes risk (MBR) decoding). Our results show that MBR decoding is a very effective method, that translation quality can be improved using a small number of samples, and that instruction tuning has a strong impact on the relation between the diversity of the hypotheses and the sampling temperature.

An Empirical Study of Translation Hypothesis Ensembling with Large Language Models

emnlp-main.733

2310.11430

Poster

https://aclanthology.org/2023.emnlp-main.734.bib

https://aclanthology.org/2023.emnlp-main.734/

@inproceedings{shin-etal-2023-fedtherapist, title = "{F}ed{T}herapist: Mental Health Monitoring with User-Generated Linguistic Expressions on Smartphones via Federated Learning", author = "Shin, Jaemin and Yoon, Hyungjun and Lee, Seungjoo and Park, Sungjoon and Liu, Yunxin and Choi, Jinho and Lee, Sung-Ju", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.734", doi = "10.18653/v1/2023.emnlp-main.734", pages = "11971--11988", abstract = "Psychiatrists diagnose mental disorders via the linguistic use of patients. Still, due to data privacy, existing passive mental health monitoring systems use alternative features such as activity, app usage, and location via mobile devices. We propose FedTherapist, a mobile mental health monitoring system that utilizes continuous speech and keyboard input in a privacy-preserving way via federated learning. We explore multiple model designs by comparing their performance and overhead for FedTherapist to overcome the complex nature of on-device language model training on smartphones. We further propose a Context-Aware Language Learning (CALL) methodology to effectively utilize smartphones{'} large and noisy text for mental health signal sensing. Our IRB-approved evaluation of the prediction of self-reported depression, stress, anxiety, and mood from 46 participants shows higher accuracy of FedTherapist compared with the performance with non-language features, achieving 0.15 AUROC improvement and 8.21{\%} MAE reduction.", }

Psychiatrists diagnose mental disorders via the linguistic use of patients. Still, due to data privacy, existing passive mental health monitoring systems use alternative features such as activity, app usage, and location via mobile devices. We propose FedTherapist, a mobile mental health monitoring system that utilizes continuous speech and keyboard input in a privacy-preserving way via federated learning. We explore multiple model designs by comparing their performance and overhead for FedTherapist to overcome the complex nature of on-device language model training on smartphones. We further propose a Context-Aware Language Learning (CALL) methodology to effectively utilize smartphones{'} large and noisy text for mental health signal sensing. Our IRB-approved evaluation of the prediction of self-reported depression, stress, anxiety, and mood from 46 participants shows higher accuracy of FedTherapist compared with the performance with non-language features, achieving 0.15 AUROC improvement and 8.21{\%} MAE reduction.

FedTherapist: Mental Health Monitoring with User-Generated Linguistic Expressions on Smartphones via Federated Learning

emnlp-main.734

2310.16538

Poster

https://aclanthology.org/2023.emnlp-main.735.bib

https://aclanthology.org/2023.emnlp-main.735/

@inproceedings{kim-etal-2023-visually, title = "Visually-Situated Natural Language Understanding with Contrastive Reading Model and Frozen Large Language Models", author = "Kim, Geewook and Lee, Hodong and Kim, Daehee and Jung, Haeji and Park, Sanghee and Kim, Yoonsik and Yun, Sangdoo and Kil, Taeho and Lee, Bado and Park, Seunghyun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.735", doi = "10.18653/v1/2023.emnlp-main.735", pages = "11989--12010", abstract = "Recent advances in Large Language Models (LLMs) have stimulated a surge of research aimed at extending their applications to the visual domain. While these models exhibit promise in generating abstract image captions and facilitating natural conversations, their performance on text-rich images still requires improvement. In this paper, we introduce Contrastive Reading Model (Cream), a novel neural architecture designed to enhance the language-image understanding capability of LLMs by capturing intricate details that are often overlooked in existing methods. Cream combines vision and auxiliary encoders, fortified by a contrastive feature alignment technique, to achieve a more effective comprehension of language information in visually situated contexts within the images. Our approach bridges the gap between vision and language understanding, paving the way for the development of more sophisticated Document Intelligence Assistants. Through rigorous evaluations across diverse visually-situated language understanding tasks that demand reasoning capabilities, we demonstrate the compelling performance of Cream, positioning it as a prominent model in the field of visual document understanding. We provide our codebase and newly-generated datasets at https://github.com/naver-ai/cream.", }

Recent advances in Large Language Models (LLMs) have stimulated a surge of research aimed at extending their applications to the visual domain. While these models exhibit promise in generating abstract image captions and facilitating natural conversations, their performance on text-rich images still requires improvement. In this paper, we introduce Contrastive Reading Model (Cream), a novel neural architecture designed to enhance the language-image understanding capability of LLMs by capturing intricate details that are often overlooked in existing methods. Cream combines vision and auxiliary encoders, fortified by a contrastive feature alignment technique, to achieve a more effective comprehension of language information in visually situated contexts within the images. Our approach bridges the gap between vision and language understanding, paving the way for the development of more sophisticated Document Intelligence Assistants. Through rigorous evaluations across diverse visually-situated language understanding tasks that demand reasoning capabilities, we demonstrate the compelling performance of Cream, positioning it as a prominent model in the field of visual document understanding. We provide our codebase and newly-generated datasets at https://github.com/naver-ai/cream.

Visually-Situated Natural Language Understanding with Contrastive Reading Model and Frozen Large Language Models

emnlp-main.735

2305.15080

https://huggingface.co/papers/2305.15080

Poster

https://aclanthology.org/2023.emnlp-main.736.bib

https://aclanthology.org/2023.emnlp-main.736/

@inproceedings{liu-etal-2023-continual, title = "Continual Learning for Multilingual Neural Machine Translation via Dual Importance-based Model Division", author = "Liu, Junpeng and Huang, Kaiyu and Yu, Hao and Li, Jiuyi and Su, Jinsong and Huang, Degen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.736", doi = "10.18653/v1/2023.emnlp-main.736", pages = "12011--12027", abstract = "A persistent goal of multilingual neural machine translation (MNMT) is to continually adapt the model to support new language pairs or improve some current language pairs without accessing the previous training data. To achieve this, the existing methods primarily focus on preventing catastrophic forgetting by making compromises between the original and new language pairs, leading to sub-optimal performance on both translation tasks. To mitigate this problem, we propose a dual importance-based model division method to divide the model parameters into two parts and separately model the translation of the original and new tasks. Specifically, we first remove the parameters that are negligible to the original tasks but essential to the new tasks to obtain a pruned model, which is responsible for the original translation tasks. Then we expand the pruned model with external parameters and fine-tune the newly added parameters with new training data. The whole fine-tuned model will be used for the new translation tasks. Experimental results show that our method can efficiently adapt the original model to various new translation tasks while retaining the performance of the original tasks. Further analyses demonstrate that our method consistently outperforms several strong baselines under different incremental translation scenarios.", }

A persistent goal of multilingual neural machine translation (MNMT) is to continually adapt the model to support new language pairs or improve some current language pairs without accessing the previous training data. To achieve this, the existing methods primarily focus on preventing catastrophic forgetting by making compromises between the original and new language pairs, leading to sub-optimal performance on both translation tasks. To mitigate this problem, we propose a dual importance-based model division method to divide the model parameters into two parts and separately model the translation of the original and new tasks. Specifically, we first remove the parameters that are negligible to the original tasks but essential to the new tasks to obtain a pruned model, which is responsible for the original translation tasks. Then we expand the pruned model with external parameters and fine-tune the newly added parameters with new training data. The whole fine-tuned model will be used for the new translation tasks. Experimental results show that our method can efficiently adapt the original model to various new translation tasks while retaining the performance of the original tasks. Further analyses demonstrate that our method consistently outperforms several strong baselines under different incremental translation scenarios.

Continual Learning for Multilingual Neural Machine Translation via Dual Importance-based Model Division

emnlp-main.736

Poster

https://aclanthology.org/2023.emnlp-main.737.bib

https://aclanthology.org/2023.emnlp-main.737/

@inproceedings{xu-etal-2023-simcse, title = "{S}im{CSE}++: Improving Contrastive Learning for Sentence Embeddings from Two Perspectives", author = "Xu, Jiahao and Shao, Wei and Chen, Lihui and Liu, Lemao", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.737", doi = "10.18653/v1/2023.emnlp-main.737", pages = "12028--12040", abstract = "This paper improves contrastive learning for sentence embeddings from two perspectives: handling dropout noise and addressing feature corruption. Specifically, for the first perspective, we identify that the dropout noise from negative pairs affects the model{'}s performance. Therefore, we propose a simple yet effective method to deal with such type of noise. Secondly, we pinpoint the rank bottleneck of current solutions to feature corruption and propose a dimension-wise contrastive learning objective to address this issue. Both proposed methods are generic and can be applied to any contrastive learning based models for sentence embeddings. Experimental results on standard benchmarks demonstrate that combining both proposed methods leads to a gain of 1.8 points compared to the strong baseline SimCSE configured with BERT base. Furthermore, applying the proposed method to DiffCSE, another strong contrastive learning based baseline, results in a gain of 1.4 points.", }

This paper improves contrastive learning for sentence embeddings from two perspectives: handling dropout noise and addressing feature corruption. Specifically, for the first perspective, we identify that the dropout noise from negative pairs affects the model{'}s performance. Therefore, we propose a simple yet effective method to deal with such type of noise. Secondly, we pinpoint the rank bottleneck of current solutions to feature corruption and propose a dimension-wise contrastive learning objective to address this issue. Both proposed methods are generic and can be applied to any contrastive learning based models for sentence embeddings. Experimental results on standard benchmarks demonstrate that combining both proposed methods leads to a gain of 1.8 points compared to the strong baseline SimCSE configured with BERT base. Furthermore, applying the proposed method to DiffCSE, another strong contrastive learning based baseline, results in a gain of 1.4 points.

SimCSE++: Improving Contrastive Learning for Sentence Embeddings from Two Perspectives

emnlp-main.737

2305.13192

Poster

https://aclanthology.org/2023.emnlp-main.738.bib

https://aclanthology.org/2023.emnlp-main.738/

@inproceedings{chen-yang-2023-unlearn, title = "Unlearn What You Want to Forget: Efficient Unlearning for {LLM}s", author = "Chen, Jiaao and Yang, Diyi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.738", doi = "10.18653/v1/2023.emnlp-main.738", pages = "12041--12052", abstract = "Large language models (LLMs) have achieved significant progress from pre-training on and memorizing a wide range of textual data, however, this process might suffer from privacy issues and violations of data protection regulations. As a result, the ability to easily remove data related to individual users from such models while not deteriorating their predictive quality after the removal becomes increasingly important. To address these issues, in this work, we propose an efficient unlearning framework that could efficiently update LLMs without having to retrain the whole model after data removals, by introducing lightweight unlearning layers learned with a selective teacher-student objective into the transformers. In addition, we introduce a fusion mechanism to effectively combine different unlearning layers that learns to forget different sets of data to handle a sequence of forgetting operations. Experiments on classification and generation tasks demonstrate the effectiveness of our proposed methods compared to the state-of-the-art baselines.", }

Large language models (LLMs) have achieved significant progress from pre-training on and memorizing a wide range of textual data, however, this process might suffer from privacy issues and violations of data protection regulations. As a result, the ability to easily remove data related to individual users from such models while not deteriorating their predictive quality after the removal becomes increasingly important. To address these issues, in this work, we propose an efficient unlearning framework that could efficiently update LLMs without having to retrain the whole model after data removals, by introducing lightweight unlearning layers learned with a selective teacher-student objective into the transformers. In addition, we introduce a fusion mechanism to effectively combine different unlearning layers that learns to forget different sets of data to handle a sequence of forgetting operations. Experiments on classification and generation tasks demonstrate the effectiveness of our proposed methods compared to the state-of-the-art baselines.

Unlearn What You Want to Forget: Efficient Unlearning for LLMs

emnlp-main.738

2310.20150

Poster

https://aclanthology.org/2023.emnlp-main.739.bib

https://aclanthology.org/2023.emnlp-main.739/

@inproceedings{cripwell-etal-2023-simplicity, title = "Simplicity Level Estimate ({SLE}): A Learned Reference-Less Metric for Sentence Simplification", author = {Cripwell, Liam and Legrand, Jo{\"e}l and Gardent, Claire}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.739", doi = "10.18653/v1/2023.emnlp-main.739", pages = "12053--12059", abstract = "Automatic evaluation for sentence simplification remains a challenging problem. Most popular evaluation metrics require multiple high-quality references {--} something not readily available for simplification {--} which makes it difficult to test performance on unseen domains. Furthermore, most existing metrics conflate simplicity with correlated attributes such as fluency or meaning preservation. We propose a new learned evaluation metric {---} SLE {---} which focuses on simplicity, outperforming almost all existing metrics in terms of correlation with human judgements.", }

Automatic evaluation for sentence simplification remains a challenging problem. Most popular evaluation metrics require multiple high-quality references {--} something not readily available for simplification {--} which makes it difficult to test performance on unseen domains. Furthermore, most existing metrics conflate simplicity with correlated attributes such as fluency or meaning preservation. We propose a new learned evaluation metric {---} SLE {---} which focuses on simplicity, outperforming almost all existing metrics in terms of correlation with human judgements.

Simplicity Level Estimate (SLE): A Learned Reference-Less Metric for Sentence Simplification

emnlp-main.739

2310.08170

Poster

https://aclanthology.org/2023.emnlp-main.740.bib

https://aclanthology.org/2023.emnlp-main.740/

@inproceedings{wu-etal-2023-precedent, title = "Precedent-Enhanced Legal Judgment Prediction with {LLM} and Domain-Model Collaboration", author = "Wu, Yiquan and Zhou, Siying and Liu, Yifei and Lu, Weiming and Liu, Xiaozhong and Zhang, Yating and Sun, Changlong and Wu, Fei and Kuang, Kun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.740", doi = "10.18653/v1/2023.emnlp-main.740", pages = "12060--12075", abstract = "Legal Judgment Prediction (LJP) has become an increasingly crucial task in Legal AI, i.e., predicting the judgment of the case in terms of case fact description. Precedents are the previous legal cases with similar facts, which are the basis for the judgment of the subsequent case in national legal systems. Thus, it is worthwhile to explore the utilization of precedents in the LJP. Recent advances in deep learning have enabled a variety of techniques to be used to solve the LJP task. These can be broken down into two categories: large language models (LLMs) and domain-specific models. LLMs are capable of interpreting and generating complex natural language, while domain models are efficient in learning task-specific information. In this paper, we propose the precedent-enhanced LJP framework (PLJP) {--} a system that leverages the strength of both LLM and domain models in the context of precedents. Specifically, the domain models are designed to provide candidate labels and find the proper precedents efficiently, and the large models will make the final prediction with an in-context precedents comprehension. Experiments on the real-world dataset demonstrate the effectiveness of our PLJP. Moreover, our work shows a promising direction for LLM and domain-model collaboration that can be generalized to other vertical domains.", }

Legal Judgment Prediction (LJP) has become an increasingly crucial task in Legal AI, i.e., predicting the judgment of the case in terms of case fact description. Precedents are the previous legal cases with similar facts, which are the basis for the judgment of the subsequent case in national legal systems. Thus, it is worthwhile to explore the utilization of precedents in the LJP. Recent advances in deep learning have enabled a variety of techniques to be used to solve the LJP task. These can be broken down into two categories: large language models (LLMs) and domain-specific models. LLMs are capable of interpreting and generating complex natural language, while domain models are efficient in learning task-specific information. In this paper, we propose the precedent-enhanced LJP framework (PLJP) {--} a system that leverages the strength of both LLM and domain models in the context of precedents. Specifically, the domain models are designed to provide candidate labels and find the proper precedents efficiently, and the large models will make the final prediction with an in-context precedents comprehension. Experiments on the real-world dataset demonstrate the effectiveness of our PLJP. Moreover, our work shows a promising direction for LLM and domain-model collaboration that can be generalized to other vertical domains.

Precedent-Enhanced Legal Judgment Prediction with LLM and Domain-Model Collaboration

emnlp-main.740

2310.09241

Oral

https://aclanthology.org/2023.emnlp-main.741.bib

https://aclanthology.org/2023.emnlp-main.741/

@inproceedings{min-etal-2023-factscore, title = "{FA}ct{S}core: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation", author = "Min, Sewon and Krishna, Kalpesh and Lyu, Xinxi and Lewis, Mike and Yih, Wen-tau and Koh, Pang and Iyyer, Mohit and Zettlemoyer, Luke and Hajishirzi, Hannaneh", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.741", doi = "10.18653/v1/2023.emnlp-main.741", pages = "12076--12100", abstract = "Evaluating the factuality of long-form text generated by large language models (LMs) is non-trivial because (1) generations often contain a mixture of supported and unsupported pieces of information, making binary judgments of quality inadequate, and (2) human evaluation is time-consuming and costly. In this paper, we introduce FACTSCORE, a new evaluation that breaks a generation into a series of atomic facts and computes the percentage of atomic facts supported by a reliable knowledge source. We conduct an extensive human evaluation to obtain FACTSCOREs of people biographies generated by several state-of-the-art commercial LMs{---}InstructGPT, ChatGPT, and the retrieval-augmented PerplexityAI{---}and report new analysis demonstrating the need for such a fine-grained score (e.g., ChatGPT only achieves 58{\%}). Since human evaluation is costly, we also introduce an automated model that estimates FACTSCORE using retrieval and a strong language model, with less than a 2{\%} error rate. Finally, we use this automated metric to evaluate 6,500 generations from a new set of 13 recent LMs that would have cost {\$}26K if evaluated by humans, with various findings: GPT-4 and ChatGPT are more factual than public models, and Vicuna and Alpaca are some of the best public models. FACTSCORE is available for public use via {`}pip install factscore{`}.", }

Evaluating the factuality of long-form text generated by large language models (LMs) is non-trivial because (1) generations often contain a mixture of supported and unsupported pieces of information, making binary judgments of quality inadequate, and (2) human evaluation is time-consuming and costly. In this paper, we introduce FACTSCORE, a new evaluation that breaks a generation into a series of atomic facts and computes the percentage of atomic facts supported by a reliable knowledge source. We conduct an extensive human evaluation to obtain FACTSCOREs of people biographies generated by several state-of-the-art commercial LMs{---}InstructGPT, ChatGPT, and the retrieval-augmented PerplexityAI{---}and report new analysis demonstrating the need for such a fine-grained score (e.g., ChatGPT only achieves 58{\%}). Since human evaluation is costly, we also introduce an automated model that estimates FACTSCORE using retrieval and a strong language model, with less than a 2{\%} error rate. Finally, we use this automated metric to evaluate 6,500 generations from a new set of 13 recent LMs that would have cost {\$}26K if evaluated by humans, with various findings: GPT-4 and ChatGPT are more factual than public models, and Vicuna and Alpaca are some of the best public models. FACTSCORE is available for public use via {`}pip install factscore{`}.

FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation

emnlp-main.741

2305.14251

https://huggingface.co/papers/2305.14251

Oral

https://aclanthology.org/2023.emnlp-main.742.bib

https://aclanthology.org/2023.emnlp-main.742/

@inproceedings{kadlcik-etal-2023-calc, title = "Calc-{X} and Calcformers: Empowering Arithmetical Chain-of-Thought through Interaction with Symbolic Systems", author = "Kadl{\v{c}}{\'\i}k, Marek and {\v{S}}tef{\'a}nik, Michal and Sotolar, Ondrej and Martinek, Vlastimil", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.742", doi = "10.18653/v1/2023.emnlp-main.742", pages = "12101--12108", abstract = "Despite outstanding performance in many tasks, language models are notoriously inclined to make factual errors in tasks requiring arithmetic computation. We address this deficiency by creating Calc-X, a collection of datasets that demonstrates the appropriate use of a calculator in reasoning chains. Calc-X is suitable for teaching language models to offload computations to a symbolic system. We survey and unify several existing chain-of-thought datasets into a proposed format, resulting in a standard collection of over 300,000 samples requiring arithmetic reasoning. Finally, we use the new Calc-X collection to train open-source calculator-using models and show that these models approximately double the accuracy of generating correct results compared to vanilla language model baselines.", }

Despite outstanding performance in many tasks, language models are notoriously inclined to make factual errors in tasks requiring arithmetic computation. We address this deficiency by creating Calc-X, a collection of datasets that demonstrates the appropriate use of a calculator in reasoning chains. Calc-X is suitable for teaching language models to offload computations to a symbolic system. We survey and unify several existing chain-of-thought datasets into a proposed format, resulting in a standard collection of over 300,000 samples requiring arithmetic reasoning. Finally, we use the new Calc-X collection to train open-source calculator-using models and show that these models approximately double the accuracy of generating correct results compared to vanilla language model baselines.

Calc-X and Calcformers: Empowering Arithmetical Chain-of-Thought through Interaction with Symbolic Systems

emnlp-main.742

2305.15017

https://huggingface.co/papers/2305.15017

Poster

https://aclanthology.org/2023.emnlp-main.743.bib

https://aclanthology.org/2023.emnlp-main.743/

@inproceedings{nguyen-etal-2023-cof, title = "{C}o{F}-{C}o{T}: Enhancing Large Language Models with Coarse-to-Fine Chain-of-Thought Prompting for Multi-domain {NLU} Tasks", author = "Nguyen, Hoang and Liu, Ye and Zhang, Chenwei and Zhang, Tao and Yu, Philip", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.743", doi = "10.18653/v1/2023.emnlp-main.743", pages = "12109--12119", abstract = "While Chain-of-Thought prompting is popular in reasoning tasks, its application to Large Language Models (LLMs) in Natural Language Understanding (NLU) is under-explored. Motivated by multi-step reasoning of LLMs, we propose Coarse-to-Fine Chain-of-Thought (CoF-CoT) approach that breaks down NLU tasks into multiple reasoning steps where LLMs can learn to acquire and leverage essential concepts to solve tasks from different granularities. Moreover, we propose leveraging semantic-based Abstract Meaning Representation (AMR) structured knowledge as an intermediate step to capture the nuances and diverse structures of utterances, and to understand connections between their varying levels of granularity. Our proposed approach is demonstrated effective in assisting the LLMs adapt to the multi-grained NLU tasks under both zero-shot and few-shot multi-domain settings.", }

While Chain-of-Thought prompting is popular in reasoning tasks, its application to Large Language Models (LLMs) in Natural Language Understanding (NLU) is under-explored. Motivated by multi-step reasoning of LLMs, we propose Coarse-to-Fine Chain-of-Thought (CoF-CoT) approach that breaks down NLU tasks into multiple reasoning steps where LLMs can learn to acquire and leverage essential concepts to solve tasks from different granularities. Moreover, we propose leveraging semantic-based Abstract Meaning Representation (AMR) structured knowledge as an intermediate step to capture the nuances and diverse structures of utterances, and to understand connections between their varying levels of granularity. Our proposed approach is demonstrated effective in assisting the LLMs adapt to the multi-grained NLU tasks under both zero-shot and few-shot multi-domain settings.

CoF-CoT: Enhancing Large Language Models with Coarse-to-Fine Chain-of-Thought Prompting for Multi-domain NLU Tasks

emnlp-main.743

2310.14623

Poster

https://aclanthology.org/2023.emnlp-main.744.bib

https://aclanthology.org/2023.emnlp-main.744/

@inproceedings{hanna-etal-2023-language, title = "When Language Models Fall in Love: {A}nimacy Processing in Transformer Language Models", author = "Hanna, Michael and Belinkov, Yonatan and Pezzelle, Sandro", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.744", doi = "10.18653/v1/2023.emnlp-main.744", pages = "12120--12135", abstract = "Animacy{---}whether an entity is alive and sentient{---}is fundamental to cognitive processing, impacting areas such as memory, vision, and language. However, animacy is not always expressed directly in language: in English it often manifests indirectly, in the form of selectional constraints on verbs and adjectives. This poses a potential issue for transformer language models (LMs): they often train only on text, and thus lack access to extralinguistic information from which humans learn about animacy. We ask: how does this impact LMs{'} animacy processing{---}do they still behave as humans do? We answer this question using open-source LMs. Like previous studies, we find that LMs behave much like humans when presented with entities whose animacy is typical. However, we also show that even when presented with stories about atypically animate entities, such as a peanut in love, LMs adapt: they treat these entities as animate, though they do not adapt as well as humans. Even when the context indicating atypical animacy is very short, LMs pick up on subtle clues and change their behavior. We conclude that despite the limited signal through which LMs can learn about animacy, they are indeed sensitive to the relevant lexical semantic nuances available in English.", }

Animacy{---}whether an entity is alive and sentient{---}is fundamental to cognitive processing, impacting areas such as memory, vision, and language. However, animacy is not always expressed directly in language: in English it often manifests indirectly, in the form of selectional constraints on verbs and adjectives. This poses a potential issue for transformer language models (LMs): they often train only on text, and thus lack access to extralinguistic information from which humans learn about animacy. We ask: how does this impact LMs{'} animacy processing{---}do they still behave as humans do? We answer this question using open-source LMs. Like previous studies, we find that LMs behave much like humans when presented with entities whose animacy is typical. However, we also show that even when presented with stories about atypically animate entities, such as a peanut in love, LMs adapt: they treat these entities as animate, though they do not adapt as well as humans. Even when the context indicating atypical animacy is very short, LMs pick up on subtle clues and change their behavior. We conclude that despite the limited signal through which LMs can learn about animacy, they are indeed sensitive to the relevant lexical semantic nuances available in English.

When Language Models Fall in Love: Animacy Processing in Transformer Language Models

emnlp-main.744

2310.15004

Poster

https://aclanthology.org/2023.emnlp-main.745.bib

https://aclanthology.org/2023.emnlp-main.745/

@inproceedings{wang-etal-2023-improving-unsupervised, title = "Improving Unsupervised Relation Extraction by Augmenting Diverse Sentence Pairs", author = "Wang, Qing and Zhou, Kang and Qiao, Qiao and Li, Yuepei and Li, Qi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.745", doi = "10.18653/v1/2023.emnlp-main.745", pages = "12136--12147", abstract = "Unsupervised relation extraction (URE) aims to extract relations between named entities from raw text without requiring manual annotations or pre-existing knowledge bases. In recent studies of URE, researchers put a notable emphasis on contrastive learning strategies for acquiring relation representations. However, these studies often overlook two important aspects: the inclusion of diverse positive pairs for contrastive learning and the exploration of appropriate loss functions. In this paper, we propose AugURE with both within-sentence pairs augmentation and augmentation through cross-sentence pairs extraction to increase the diversity of positive pairs and strengthen the discriminative power of contrastive learning. We also identify the limitation of noise-contrastive estimation (NCE) loss for relation representation learning and propose to apply margin loss for sentence pairs. Experiments on NYT-FB and TACRED datasets demonstrate that the proposed relation representation learning and a simple K-Means clustering achieves state-of-the-art performance.", }

Unsupervised relation extraction (URE) aims to extract relations between named entities from raw text without requiring manual annotations or pre-existing knowledge bases. In recent studies of URE, researchers put a notable emphasis on contrastive learning strategies for acquiring relation representations. However, these studies often overlook two important aspects: the inclusion of diverse positive pairs for contrastive learning and the exploration of appropriate loss functions. In this paper, we propose AugURE with both within-sentence pairs augmentation and augmentation through cross-sentence pairs extraction to increase the diversity of positive pairs and strengthen the discriminative power of contrastive learning. We also identify the limitation of noise-contrastive estimation (NCE) loss for relation representation learning and propose to apply margin loss for sentence pairs. Experiments on NYT-FB and TACRED datasets demonstrate that the proposed relation representation learning and a simple K-Means clustering achieves state-of-the-art performance.

Improving Unsupervised Relation Extraction by Augmenting Diverse Sentence Pairs

emnlp-main.745

2312.00552

Poster

https://aclanthology.org/2023.emnlp-main.746.bib

https://aclanthology.org/2023.emnlp-main.746/

@inproceedings{wahle-etal-2023-paraphrase, title = "Paraphrase Types for Generation and Detection", author = "Wahle, Jan Philip and Gipp, Bela and Ruas, Terry", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.746", doi = "10.18653/v1/2023.emnlp-main.746", pages = "12148--12164", abstract = "Current approaches in paraphrase generation and detection heavily rely on a single general similarity score, ignoring the intricate linguistic properties of language. This paper introduces two new tasks to address this shortcoming by considering paraphrase types - specific linguistic perturbations at particular text positions. We name these tasks Paraphrase Type Generation and Paraphrase Type Detection. Our results suggest that while current techniques perform well in a binary classification scenario, i.e., paraphrased or not, the inclusion of fine-grained paraphrase types poses a significant challenge. While most approaches are good at generating and detecting general semantic similar content, they fail to understand the intrinsic linguistic variables they manipulate. Models trained in generating and identifying paraphrase types also show improvements in tasks without them. In addition, scaling these models further improves their ability to understand paraphrase types. We believe paraphrase types can unlock a new paradigm for developing paraphrase models and solving tasks in the future.", }

Current approaches in paraphrase generation and detection heavily rely on a single general similarity score, ignoring the intricate linguistic properties of language. This paper introduces two new tasks to address this shortcoming by considering paraphrase types - specific linguistic perturbations at particular text positions. We name these tasks Paraphrase Type Generation and Paraphrase Type Detection. Our results suggest that while current techniques perform well in a binary classification scenario, i.e., paraphrased or not, the inclusion of fine-grained paraphrase types poses a significant challenge. While most approaches are good at generating and detecting general semantic similar content, they fail to understand the intrinsic linguistic variables they manipulate. Models trained in generating and identifying paraphrase types also show improvements in tasks without them. In addition, scaling these models further improves their ability to understand paraphrase types. We believe paraphrase types can unlock a new paradigm for developing paraphrase models and solving tasks in the future.

Paraphrase Types for Generation and Detection

emnlp-main.746

2310.14863

Poster

https://aclanthology.org/2023.emnlp-main.747.bib

https://aclanthology.org/2023.emnlp-main.747/

@inproceedings{zhang-etal-2023-target-source, title = "Target-to-Source Augmentation for Aspect Sentiment Triplet Extraction", author = "Zhang, Yice and Yang, Yifan and Li, Meng and Liang, Bin and Chen, Shiwei and Xu, Ruifeng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.747", doi = "10.18653/v1/2023.emnlp-main.747", pages = "12165--12177", abstract = "Aspect Sentiment Triplet Extraction (ASTE) is an important task in sentiment analysis, aiming to extract aspect-level opinions and sentiments from user-generated reviews. The fine-grained nature of ASTE incurs a high annotation cost, while the scarcity of annotated data limits the performance of existing methods. This paper exploits data augmentation to address this issue. Traditional augmentation methods typically modify the input sentences of existing samples via heuristic rules or language models, which have shown success in text classification tasks. However, applying these methods to fine-grained tasks like ASTE poses challenges in generating diverse augmented samples while maintaining alignment between modified sentences and origin labels. Therefore, this paper proposes a target-to-source augmentation approach for ASTE. Our approach focuses on learning a generator that can directly generate new sentences based on labels and syntactic templates. With this generator, we can generate a substantial number of diverse augmented samples by mixing labels and syntactic templates from different samples. Besides, to ensure the quality of the generated sentence, we introduce fluency and alignment discriminators to provide feedback on the generated sentence and then use this feedback to optimize the generator via a reinforcement learning framework. Experiments demonstrate that our approach significantly enhances the performance of existing ASTE models.", }

Aspect Sentiment Triplet Extraction (ASTE) is an important task in sentiment analysis, aiming to extract aspect-level opinions and sentiments from user-generated reviews. The fine-grained nature of ASTE incurs a high annotation cost, while the scarcity of annotated data limits the performance of existing methods. This paper exploits data augmentation to address this issue. Traditional augmentation methods typically modify the input sentences of existing samples via heuristic rules or language models, which have shown success in text classification tasks. However, applying these methods to fine-grained tasks like ASTE poses challenges in generating diverse augmented samples while maintaining alignment between modified sentences and origin labels. Therefore, this paper proposes a target-to-source augmentation approach for ASTE. Our approach focuses on learning a generator that can directly generate new sentences based on labels and syntactic templates. With this generator, we can generate a substantial number of diverse augmented samples by mixing labels and syntactic templates from different samples. Besides, to ensure the quality of the generated sentence, we introduce fluency and alignment discriminators to provide feedback on the generated sentence and then use this feedback to optimize the generator via a reinforcement learning framework. Experiments demonstrate that our approach significantly enhances the performance of existing ASTE models.

Target-to-Source Augmentation for Aspect Sentiment Triplet Extraction

emnlp-main.747

Poster

https://aclanthology.org/2023.emnlp-main.748.bib

https://aclanthology.org/2023.emnlp-main.748/

@inproceedings{liu-etal-2023-pac, title = "{PAC}-tuning: Fine-tuning Pre-trained Language Models with {PAC}-driven Perturbed Gradient Descent", author = "Liu, Guangliang and Xue, Zhiyu and Zhang, Xitong and Johnson, Kristen and Wang, Rongrong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.748", doi = "10.18653/v1/2023.emnlp-main.748", pages = "12178--12189", abstract = "Fine-tuning pretrained language models (PLMs) for downstream tasks is a large-scale optimization problem, in which the choice of the training algorithm critically determines how well the trained model can generalize to unseen test data, especially in the context of few-shot learning. To achieve good generalization performance and avoid overfitting, techniques such as data augmentation and pruning are often applied. However, adding these regularizations necessitates heavy tuning of the hyperparameters of optimization algorithms, such as the popular Adam optimizer. In this paper, we propose a two-stage fine-tuning method, PAC-tuning, to address this optimization challenge. First, based on PAC-Bayes training, PAC-tuning directly minimizes the PAC-Bayes generalization bound to learn proper parameter distribution. Second, PAC-tuning modifies the gradient by injecting noise with the variance learned in the first stage into the model parameters during training, resulting in a variant of perturbed gradient descent (PGD). In the past, the few-shot scenario posed difficulties for PAC-Bayes training because the PAC-Bayes bound, when applied to large models with limited training data, might not be stringent. Our experimental results across 5 GLUE benchmark tasks demonstrate that PAC-tuning successfully handles the challenges of fine-tuning tasks and outperforms strong baseline methods by a visible margin, further confirming the potential to apply PAC training for any other settings where the Adam optimizer is currently used for training.", }

Fine-tuning pretrained language models (PLMs) for downstream tasks is a large-scale optimization problem, in which the choice of the training algorithm critically determines how well the trained model can generalize to unseen test data, especially in the context of few-shot learning. To achieve good generalization performance and avoid overfitting, techniques such as data augmentation and pruning are often applied. However, adding these regularizations necessitates heavy tuning of the hyperparameters of optimization algorithms, such as the popular Adam optimizer. In this paper, we propose a two-stage fine-tuning method, PAC-tuning, to address this optimization challenge. First, based on PAC-Bayes training, PAC-tuning directly minimizes the PAC-Bayes generalization bound to learn proper parameter distribution. Second, PAC-tuning modifies the gradient by injecting noise with the variance learned in the first stage into the model parameters during training, resulting in a variant of perturbed gradient descent (PGD). In the past, the few-shot scenario posed difficulties for PAC-Bayes training because the PAC-Bayes bound, when applied to large models with limited training data, might not be stringent. Our experimental results across 5 GLUE benchmark tasks demonstrate that PAC-tuning successfully handles the challenges of fine-tuning tasks and outperforms strong baseline methods by a visible margin, further confirming the potential to apply PAC training for any other settings where the Adam optimizer is currently used for training.

PAC-tuning: Fine-tuning Pre-trained Language Models with PAC-driven Perturbed Gradient Descent

emnlp-main.748

Poster

https://aclanthology.org/2023.emnlp-main.749.bib

https://aclanthology.org/2023.emnlp-main.749/

@inproceedings{yun-etal-2023-emergence, title = "Emergence of Abstract State Representations in Embodied Sequence Modeling", author = "Yun, Tian and Zeng, Zilai and Handa, Kunal and Thapliyal, Ashish and Pang, Bo and Pavlick, Ellie and Sun, Chen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.749", doi = "10.18653/v1/2023.emnlp-main.749", pages = "12190--12205", abstract = "Decision making via sequence modeling aims to mimic the success of language models, where actions taken by an embodied agent are modeled as tokens to predict. Despite their promising performance, it remains unclear if embodied sequence modeling leads to the emergence of internal representations that represent the environmental state information. A model that lacks abstract state representations would be liable to make decisions based on surface statistics which fail to generalize. We take the BabyAI environment, a grid world in which language-conditioned navigation tasks are performed, and build a sequence modeling Transformer, which takes a language instruction, a sequence of actions, and environmental observations as its inputs. In order to investigate the emergence of abstract state representations, we design a {``}blindfolded{''} navigation task, where only the initial environmental layout, the language instruction, and the action sequence to complete the task are available for training. Our probing results show that intermediate environmental layouts can be reasonably reconstructed from the internal activations of a trained model, and that language instructions play a role in the reconstruction accuracy. Our results suggest that many key features of state representations can emerge via embodied sequence modeling, supporting an optimistic outlook for applications of sequence modeling objectives to more complex embodied decision-making domains.", }

Decision making via sequence modeling aims to mimic the success of language models, where actions taken by an embodied agent are modeled as tokens to predict. Despite their promising performance, it remains unclear if embodied sequence modeling leads to the emergence of internal representations that represent the environmental state information. A model that lacks abstract state representations would be liable to make decisions based on surface statistics which fail to generalize. We take the BabyAI environment, a grid world in which language-conditioned navigation tasks are performed, and build a sequence modeling Transformer, which takes a language instruction, a sequence of actions, and environmental observations as its inputs. In order to investigate the emergence of abstract state representations, we design a {``}blindfolded{''} navigation task, where only the initial environmental layout, the language instruction, and the action sequence to complete the task are available for training. Our probing results show that intermediate environmental layouts can be reasonably reconstructed from the internal activations of a trained model, and that language instructions play a role in the reconstruction accuracy. Our results suggest that many key features of state representations can emerge via embodied sequence modeling, supporting an optimistic outlook for applications of sequence modeling objectives to more complex embodied decision-making domains.

Emergence of Abstract State Representations in Embodied Sequence Modeling

emnlp-main.749

2311.02171

Poster

https://aclanthology.org/2023.emnlp-main.750.bib

https://aclanthology.org/2023.emnlp-main.750/

@inproceedings{qin-zhong-2023-accelerating, title = "Accelerating Toeplitz Neural Network with Constant-time Inference Complexity", author = "Qin, Zhen and Zhong, Yiran", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.750", doi = "10.18653/v1/2023.emnlp-main.750", pages = "12206--12215", abstract = "Toeplitz Neural Networks (TNNs) have exhibited outstanding performance in various sequence modeling tasks. They outperform commonly used Transformer-based models while benefiting from log-linear space-time complexities. On the other hand, State Space Models (SSMs) achieve lower performance than TNNs in language modeling but offer the advantage of constant inference complexity. In this paper, we aim to combine the strengths of TNNs and SSMs by converting TNNs to SSMs during inference, thereby enabling TNNs to achieve the same constant inference complexities as SSMs. To accomplish this, we formulate the conversion process as an optimization problem and provide a closed-form solution. We demonstrate how to transform the target equation into a Vandermonde linear system problem, which can be efficiently solved using the Discrete Fourier Transform (DFT). Notably, our method requires no training and maintains numerical stability. It can be also applied to any LongConv-based model. To assess its effectiveness, we conduct extensive experiments on language modeling tasks across various settings. Additionally, we compare our method to other gradient-descent solutions, highlighting the superior numerical stability of our approach. The source code is available at https://github.com/OpenNLPLab/ETSC-Exact-Toeplitz-to-SSM-Conversion.", }

Toeplitz Neural Networks (TNNs) have exhibited outstanding performance in various sequence modeling tasks. They outperform commonly used Transformer-based models while benefiting from log-linear space-time complexities. On the other hand, State Space Models (SSMs) achieve lower performance than TNNs in language modeling but offer the advantage of constant inference complexity. In this paper, we aim to combine the strengths of TNNs and SSMs by converting TNNs to SSMs during inference, thereby enabling TNNs to achieve the same constant inference complexities as SSMs. To accomplish this, we formulate the conversion process as an optimization problem and provide a closed-form solution. We demonstrate how to transform the target equation into a Vandermonde linear system problem, which can be efficiently solved using the Discrete Fourier Transform (DFT). Notably, our method requires no training and maintains numerical stability. It can be also applied to any LongConv-based model. To assess its effectiveness, we conduct extensive experiments on language modeling tasks across various settings. Additionally, we compare our method to other gradient-descent solutions, highlighting the superior numerical stability of our approach. The source code is available at https://github.com/OpenNLPLab/ETSC-Exact-Toeplitz-to-SSM-Conversion.

Accelerating Toeplitz Neural Network with Constant-time Inference Complexity

emnlp-main.750

2311.08756

https://huggingface.co/papers/2311.08756

Poster

https://aclanthology.org/2023.emnlp-main.751.bib

https://aclanthology.org/2023.emnlp-main.751/

@inproceedings{geva-etal-2023-dissecting, title = "Dissecting Recall of Factual Associations in Auto-Regressive Language Models", author = "Geva, Mor and Bastings, Jasmijn and Filippova, Katja and Globerson, Amir", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.751", doi = "10.18653/v1/2023.emnlp-main.751", pages = "12216--12235", abstract = "Transformer-based language models (LMs) are known to capture factual knowledge in their parameters. While previous work looked into where factual associations are stored, only little is known about how they are retrieved internally during inference. We investigate this question through the lens of information flow. Given a subject-relation query, we study how the model aggregates information about the subject and relation to predict the correct attribute. With interventions on attention edges, we first identify two critical points where information propagates to the prediction: one from the relation positions followed by another from the subject positions. Next, by analyzing the information at these points, we unveil a three-step internal mechanism for attribute extraction. First, the representation at the last-subject position goes through an enrichment process, driven by the early MLP sublayers, to encode many subject-related attributes. Second, information from the relation propagates to the prediction. Third, the prediction representation {``}queries{''} the enriched subject to extract the attribute. Perhaps surprisingly, this extraction is typically done via attention heads, which often encode subject-attribute mappings in their parameters. Overall, our findings introduce a comprehensive view of how factual associations are stored and extracted internally in LMs, facilitating future research on knowledge localization and editing.", }

Transformer-based language models (LMs) are known to capture factual knowledge in their parameters. While previous work looked into where factual associations are stored, only little is known about how they are retrieved internally during inference. We investigate this question through the lens of information flow. Given a subject-relation query, we study how the model aggregates information about the subject and relation to predict the correct attribute. With interventions on attention edges, we first identify two critical points where information propagates to the prediction: one from the relation positions followed by another from the subject positions. Next, by analyzing the information at these points, we unveil a three-step internal mechanism for attribute extraction. First, the representation at the last-subject position goes through an enrichment process, driven by the early MLP sublayers, to encode many subject-related attributes. Second, information from the relation propagates to the prediction. Third, the prediction representation {``}queries{''} the enriched subject to extract the attribute. Perhaps surprisingly, this extraction is typically done via attention heads, which often encode subject-attribute mappings in their parameters. Overall, our findings introduce a comprehensive view of how factual associations are stored and extracted internally in LMs, facilitating future research on knowledge localization and editing.

Dissecting Recall of Factual Associations in Auto-Regressive Language Models

emnlp-main.751

2304.14767

Oral

https://aclanthology.org/2023.emnlp-main.752.bib

https://aclanthology.org/2023.emnlp-main.752/

@inproceedings{jeoung-etal-2023-stereomap, title = "{S}tereo{M}ap: Quantifying the Awareness of Human-like Stereotypes in Large Language Models", author = "Jeoung, Sullam and Ge, Yubin and Diesner, Jana", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.752", doi = "10.18653/v1/2023.emnlp-main.752", pages = "12236--12256", abstract = "Large Language Models (LLMs) have been observed to encode and perpetuate harmful associations present in the training data. We propose a theoretically grounded framework called StereoMap to gain insights into their perceptions of how demographic groups have been viewed by society. The framework is grounded in the Stereotype Content Model (SCM); a well-established theory from psychology. According to SCM, stereotypes are not all alike. Instead, the dimensions of Warmth and Competence serve as the factors that delineate the nature of stereotypes. Based on the SCM theory, StereoMap maps LLMs{'} perceptions of social groups (defined by socio-demographic features) using the dimensions of Warmth and Competence. Furthermore, the framework enables the investigation of keywords and verbalizations of reasoning of LLMs{'} judgments to uncover underlying factors influencing their perceptions. Our results show that LLMs exhibit a diverse range of perceptions towards these groups, characterized by mixed evaluations along the dimensions of Warmth and Competence. Furthermore, analyzing the reasonings of LLMs, our findings indicate that LLMs demonstrate an awareness of social disparities, often stating statistical data and research findings to support their reasoning. This study contributes to the understanding of how LLMs perceive and represent social groups, shedding light on their potential biases and the perpetuation of harmful associations.", }

Large Language Models (LLMs) have been observed to encode and perpetuate harmful associations present in the training data. We propose a theoretically grounded framework called StereoMap to gain insights into their perceptions of how demographic groups have been viewed by society. The framework is grounded in the Stereotype Content Model (SCM); a well-established theory from psychology. According to SCM, stereotypes are not all alike. Instead, the dimensions of Warmth and Competence serve as the factors that delineate the nature of stereotypes. Based on the SCM theory, StereoMap maps LLMs{'} perceptions of social groups (defined by socio-demographic features) using the dimensions of Warmth and Competence. Furthermore, the framework enables the investigation of keywords and verbalizations of reasoning of LLMs{'} judgments to uncover underlying factors influencing their perceptions. Our results show that LLMs exhibit a diverse range of perceptions towards these groups, characterized by mixed evaluations along the dimensions of Warmth and Competence. Furthermore, analyzing the reasonings of LLMs, our findings indicate that LLMs demonstrate an awareness of social disparities, often stating statistical data and research findings to support their reasoning. This study contributes to the understanding of how LLMs perceive and represent social groups, shedding light on their potential biases and the perpetuation of harmful associations.

StereoMap: Quantifying the Awareness of Human-like Stereotypes in Large Language Models

emnlp-main.752

2310.13673

Poster

https://aclanthology.org/2023.emnlp-main.753.bib

https://aclanthology.org/2023.emnlp-main.753/

@inproceedings{pham-etal-2023-select, title = "Select, Prompt, Filter: Distilling Large Language Models for Summarizing Conversations", author = "Pham, Minh-Quang and Indurthi, Sathish and Chollampatt, Shamil and Turchi, Marco", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.753", doi = "10.18653/v1/2023.emnlp-main.753", pages = "12257--12265", abstract = "Large language models (LLMs) like ChatGPT can be expensive to train, deploy, and use for specific natural language generation tasks such as text summarization and for certain domains. A promising alternative is to fine-tune relatively smaller language models (LMs) on a particular task using high-quality, in-domain datasets. However, it can be prohibitively expensive to get such high-quality training data. This issue has been mitigated by generating weakly supervised data via knowledge distillation (KD) of LLMs. We propose a three-step approach to distill ChatGPT and fine-tune smaller LMs for summarizing forum conversations. More specifically, we design a method to selectively sample a large unannotated corpus of forum conversation using a semantic similarity metric. Then, we use the same metric to retrieve suitable prompts for ChatGPT from a small annotated validation set in the same domain. The generated dataset is then filtered to remove low-quality instances. Our proposed select-prompt-filter KD approach leads to significant improvements of up to 6.6 ROUGE-2 score by leveraging sufficient in-domain pseudo-labeled data over a standard KD approach given the same size of training data.", }

Large language models (LLMs) like ChatGPT can be expensive to train, deploy, and use for specific natural language generation tasks such as text summarization and for certain domains. A promising alternative is to fine-tune relatively smaller language models (LMs) on a particular task using high-quality, in-domain datasets. However, it can be prohibitively expensive to get such high-quality training data. This issue has been mitigated by generating weakly supervised data via knowledge distillation (KD) of LLMs. We propose a three-step approach to distill ChatGPT and fine-tune smaller LMs for summarizing forum conversations. More specifically, we design a method to selectively sample a large unannotated corpus of forum conversation using a semantic similarity metric. Then, we use the same metric to retrieve suitable prompts for ChatGPT from a small annotated validation set in the same domain. The generated dataset is then filtered to remove low-quality instances. Our proposed select-prompt-filter KD approach leads to significant improvements of up to 6.6 ROUGE-2 score by leveraging sufficient in-domain pseudo-labeled data over a standard KD approach given the same size of training data.

Select, Prompt, Filter: Distilling Large Language Models for Summarizing Conversations

emnlp-main.753

Poster

https://aclanthology.org/2023.emnlp-main.754.bib

https://aclanthology.org/2023.emnlp-main.754/

@inproceedings{wein-2023-human, title = "Human Raters Cannot Distinguish {E}nglish Translations from Original {E}nglish Texts", author = "Wein, Shira", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.754", doi = "10.18653/v1/2023.emnlp-main.754", pages = "12266--12272", abstract = "The term translationese describes the set of linguistic features unique to translated texts, which appear regardless of translation quality. Though automatic classifiers designed to distinguish translated texts achieve high accuracy and prior work has identified common hallmarks of translationese, human accuracy of identifying translated text is understudied. In this work, we perform a human evaluation of English original/translated texts in order to explore raters{'} ability to classify texts as being original or translated English and the features that lead a rater to judge text as being translated. Ultimately, we find that, regardless of the annotators{'} native language or the source language of the text, annotators are unable to distinguish translations from original English texts and also have low agreement. Our results provide critical insight into work in translation studies and context for assessments of translationese classifiers.", }

The term translationese describes the set of linguistic features unique to translated texts, which appear regardless of translation quality. Though automatic classifiers designed to distinguish translated texts achieve high accuracy and prior work has identified common hallmarks of translationese, human accuracy of identifying translated text is understudied. In this work, we perform a human evaluation of English original/translated texts in order to explore raters{'} ability to classify texts as being original or translated English and the features that lead a rater to judge text as being translated. Ultimately, we find that, regardless of the annotators{'} native language or the source language of the text, annotators are unable to distinguish translations from original English texts and also have low agreement. Our results provide critical insight into work in translation studies and context for assessments of translationese classifiers.

Human Raters Cannot Distinguish English Translations from Original English Texts

emnlp-main.754

Poster

https://aclanthology.org/2023.emnlp-main.755.bib

https://aclanthology.org/2023.emnlp-main.755/

@inproceedings{kruk-etal-2023-impressions, title = "Impressions: Visual Semiotics and Aesthetic Impact Understanding", author = "Kruk, Julia and Ziems, Caleb and Yang, Diyi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.755", doi = "10.18653/v1/2023.emnlp-main.755", pages = "12273--12291", abstract = "Is aesthetic impact different from beauty? Is visual salience a reflection of its capacity for effective communication? We present Impressions, a novel dataset through which to investigate the semiotics of images, and how specific visual features and design choices can elicit specific emotions, thoughts and beliefs. We posit that the impactfulness of an image extends beyond formal definitions of aesthetics, to its success as a communicative act, where style contributes as much to meaning formation as the subject matter. We also acknowledge that existing Image Captioning datasets are not designed to empower state-of-the-art architectures to model potential human impressions or interpretations of images. To fill this need, we design an annotation task heavily inspired by image analysis techniques in the Visual Arts to collect 1,440 image-caption pairs and 4,320 unique annotations exploring impact, pragmatic image description, impressions and aesthetic design choices. We show that existing multimodal image captioning and conditional generation models struggle to simulate plausible human responses to images. However, this dataset significantly improves their ability to model impressions and aesthetic evaluations of images through fine-tuning and few-shot adaptation.", }

Is aesthetic impact different from beauty? Is visual salience a reflection of its capacity for effective communication? We present Impressions, a novel dataset through which to investigate the semiotics of images, and how specific visual features and design choices can elicit specific emotions, thoughts and beliefs. We posit that the impactfulness of an image extends beyond formal definitions of aesthetics, to its success as a communicative act, where style contributes as much to meaning formation as the subject matter. We also acknowledge that existing Image Captioning datasets are not designed to empower state-of-the-art architectures to model potential human impressions or interpretations of images. To fill this need, we design an annotation task heavily inspired by image analysis techniques in the Visual Arts to collect 1,440 image-caption pairs and 4,320 unique annotations exploring impact, pragmatic image description, impressions and aesthetic design choices. We show that existing multimodal image captioning and conditional generation models struggle to simulate plausible human responses to images. However, this dataset significantly improves their ability to model impressions and aesthetic evaluations of images through fine-tuning and few-shot adaptation.

Impressions: Visual Semiotics and Aesthetic Impact Understanding

emnlp-main.755

2310.17887

https://huggingface.co/papers/2310.17887

Poster

https://aclanthology.org/2023.emnlp-main.756.bib

https://aclanthology.org/2023.emnlp-main.756/

@inproceedings{an-etal-2023-dna, title = "{DNA}: Denoised Neighborhood Aggregation for Fine-grained Category Discovery", author = "An, Wenbin and Tian, Feng and Shi, Wenkai and Chen, Yan and Zheng, Qinghua and Wang, QianYing and Chen, Ping", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.756", doi = "10.18653/v1/2023.emnlp-main.756", pages = "12292--12302", abstract = "Discovering fine-grained categories from coarsely labeled data is a practical and challenging task, which can bridge the gap between the demand for fine-grained analysis and the high annotation cost. Previous works mainly focus on instance-level discrimination to learn low-level features, but ignore semantic similarities between data, which may prevent these models learning compact cluster representations. In this paper, we propose $\textit{Denoised Neighborhood Aggregation}$ (DNA), a self-supervised framework that encodes semantic structures of data into the embedding space. Specifically, we retrieve $\textit{k}$-nearest neighbors of a query as its positive keys to capture semantic similarities between data and then aggregate information from the neighbors to learn compact cluster representations, which can make fine-grained categories more separatable. However, the retrieved neighbors can be noisy and contain many false-positive keys, which can degrade the quality of learned embeddings. To cope with this challenge, we propose three principles to filter out these false neighbors for better representation learning. Furthermore, we theoretically justify that the learning objective of our framework is equivalent to a clustering loss, which can capture semantic similarities between data to form compact fine-grained clusters. Extensive experiments on three benchmark datasets show that our method can retrieve more accurate neighbors (21.31{\%} accuracy improvement) and outperform state-of-the-art models by a large margin (average 9.96{\%} improvement on three metrics). Our code and data are available at https://github.com/Lackel/DNA.", }

Discovering fine-grained categories from coarsely labeled data is a practical and challenging task, which can bridge the gap between the demand for fine-grained analysis and the high annotation cost. Previous works mainly focus on instance-level discrimination to learn low-level features, but ignore semantic similarities between data, which may prevent these models learning compact cluster representations. In this paper, we propose $\textit{Denoised Neighborhood Aggregation}$ (DNA), a self-supervised framework that encodes semantic structures of data into the embedding space. Specifically, we retrieve $\textit{k}$-nearest neighbors of a query as its positive keys to capture semantic similarities between data and then aggregate information from the neighbors to learn compact cluster representations, which can make fine-grained categories more separatable. However, the retrieved neighbors can be noisy and contain many false-positive keys, which can degrade the quality of learned embeddings. To cope with this challenge, we propose three principles to filter out these false neighbors for better representation learning. Furthermore, we theoretically justify that the learning objective of our framework is equivalent to a clustering loss, which can capture semantic similarities between data to form compact fine-grained clusters. Extensive experiments on three benchmark datasets show that our method can retrieve more accurate neighbors (21.31{\%} accuracy improvement) and outperform state-of-the-art models by a large margin (average 9.96{\%} improvement on three metrics). Our code and data are available at https://github.com/Lackel/DNA.

DNA: Denoised Neighborhood Aggregation for Fine-grained Category Discovery

emnlp-main.756

2310.10151

Poster

https://aclanthology.org/2023.emnlp-main.757.bib

https://aclanthology.org/2023.emnlp-main.757/

@inproceedings{zhao-etal-2023-prompt, title = "Prompt as Triggers for Backdoor Attack: Examining the Vulnerability in Language Models", author = "Zhao, Shuai and Wen, Jinming and Luu, Anh and Zhao, Junbo and Fu, Jie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.757", doi = "10.18653/v1/2023.emnlp-main.757", pages = "12303--12317", abstract = "The prompt-based learning paradigm, which bridges the gap between pre-training and fine-tuning, achieves state-of-the-art performance on several NLP tasks, particularly in few-shot settings. Despite being widely applied, prompt-based learning is vulnerable to backdoor attacks. Textual backdoor attacks are designed to introduce targeted vulnerabilities into models by poisoning a subset of training samples through trigger injection and label modification. However, they suffer from flaws such as abnormal natural language expressions resulting from the trigger and incorrect labeling of poisoned samples. In this study, we propose ProAttack, a novel and efficient method for performing clean-label backdoor attacks based on the prompt, which uses the prompt itself as a trigger. Our method does not require external triggers and ensures correct labeling of poisoned samples, improving the stealthy nature of the backdoor attack. With extensive experiments on rich-resource and few-shot text classification tasks, we empirically validate ProAttack{'}s competitive performance in textual backdoor attacks. Notably, in the rich-resource setting, ProAttack achieves state-of-the-art attack success rates in the clean-label backdoor attack benchmark without external triggers.", }

The prompt-based learning paradigm, which bridges the gap between pre-training and fine-tuning, achieves state-of-the-art performance on several NLP tasks, particularly in few-shot settings. Despite being widely applied, prompt-based learning is vulnerable to backdoor attacks. Textual backdoor attacks are designed to introduce targeted vulnerabilities into models by poisoning a subset of training samples through trigger injection and label modification. However, they suffer from flaws such as abnormal natural language expressions resulting from the trigger and incorrect labeling of poisoned samples. In this study, we propose ProAttack, a novel and efficient method for performing clean-label backdoor attacks based on the prompt, which uses the prompt itself as a trigger. Our method does not require external triggers and ensures correct labeling of poisoned samples, improving the stealthy nature of the backdoor attack. With extensive experiments on rich-resource and few-shot text classification tasks, we empirically validate ProAttack{'}s competitive performance in textual backdoor attacks. Notably, in the rich-resource setting, ProAttack achieves state-of-the-art attack success rates in the clean-label backdoor attack benchmark without external triggers.

Prompt as Triggers for Backdoor Attack: Examining the Vulnerability in Language Models

emnlp-main.757

2305.01219

Poster

https://aclanthology.org/2023.emnlp-main.758.bib

https://aclanthology.org/2023.emnlp-main.758/

@inproceedings{cheng-etal-2023-uprise, title = "{UPRISE}: Universal Prompt Retrieval for Improving Zero-Shot Evaluation", author = "Cheng, Daixuan and Huang, Shaohan and Bi, Junyu and Zhan, Yuefeng and Liu, Jianfeng and Wang, Yujing and Sun, Hao and Wei, Furu and Deng, Weiwei and Zhang, Qi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.758", doi = "10.18653/v1/2023.emnlp-main.758", pages = "12318--12337", abstract = "Large Language Models (LLMs) are popular for their impressive abilities, but the need for model-specific fine-tuning or task-specific prompt engineering can hinder their generalization. We propose UPRISE (Universal Prompt Retrieval for Improving zero-Shot Evaluation), which tunes a lightweight and versatile retriever that automatically retrieves prompts for a given zero-shot task input. Specifically, we demonstrate universality in a cross-task and cross-model scenario: the retriever is tuned on diverse tasks, but tested on unseen task types; we use a small frozen LLM, GPT-Neo-2.7B, for tuning the retriever, but test the retriever on different LLMs of much larger scales, such as BLOOM-7.1B, OPT-66B and GPT3-175B. Additionally, we show that UPRISE mitigates the hallucination problem in our experiments with ChatGPT, suggesting its potential to improve even the strongest LLMs. Our model and code are available at https://github.com/microsoft/LMOps.", }

Large Language Models (LLMs) are popular for their impressive abilities, but the need for model-specific fine-tuning or task-specific prompt engineering can hinder their generalization. We propose UPRISE (Universal Prompt Retrieval for Improving zero-Shot Evaluation), which tunes a lightweight and versatile retriever that automatically retrieves prompts for a given zero-shot task input. Specifically, we demonstrate universality in a cross-task and cross-model scenario: the retriever is tuned on diverse tasks, but tested on unseen task types; we use a small frozen LLM, GPT-Neo-2.7B, for tuning the retriever, but test the retriever on different LLMs of much larger scales, such as BLOOM-7.1B, OPT-66B and GPT3-175B. Additionally, we show that UPRISE mitigates the hallucination problem in our experiments with ChatGPT, suggesting its potential to improve even the strongest LLMs. Our model and code are available at https://github.com/microsoft/LMOps.

UPRISE: Universal Prompt Retrieval for Improving Zero-Shot Evaluation

emnlp-main.758

2303.08518

https://huggingface.co/papers/2303.08518

Poster

https://aclanthology.org/2023.emnlp-main.759.bib

https://aclanthology.org/2023.emnlp-main.759/

@inproceedings{yu-etal-2023-krls, title = "{KRLS}: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning", author = "Yu, Xiao and Wu, Qingyang and Qian, Kun and Yu, Zhou", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.759", doi = "10.18653/v1/2023.emnlp-main.759", pages = "12338--12358", abstract = "In task-oriented dialogs (TOD), reinforcement learning (RL) algorithms train a model to directly optimize response for task-related metrics. However, RL often needs to perform exploration, which can be time-consuming due to the slow auto-regressive sequence generation process. We investigate an approach to create a more efficient RL-based algorithm to improve TOD performance in an offline setting. First, we use a faster generation procedure that samples from independent next-word distributions after training the language model (LM) with supervised learning. We then introduce a fine-grained reward function to help the model focus on learning key information in a dialog, by measuring the importance and semantic closeness of each generated token. Experiments on the MultiWoZ dataset show our new training algorithm, Keywords Reinforcement Learning with Next-word Sampling (KRLS), achieves state-of-the-art performance on the end-to-end response generation task, with a 15{\%} training time reduction compared to a standard RL algorithm using auto-regressive generation.", }

In task-oriented dialogs (TOD), reinforcement learning (RL) algorithms train a model to directly optimize response for task-related metrics. However, RL often needs to perform exploration, which can be time-consuming due to the slow auto-regressive sequence generation process. We investigate an approach to create a more efficient RL-based algorithm to improve TOD performance in an offline setting. First, we use a faster generation procedure that samples from independent next-word distributions after training the language model (LM) with supervised learning. We then introduce a fine-grained reward function to help the model focus on learning key information in a dialog, by measuring the importance and semantic closeness of each generated token. Experiments on the MultiWoZ dataset show our new training algorithm, Keywords Reinforcement Learning with Next-word Sampling (KRLS), achieves state-of-the-art performance on the end-to-end response generation task, with a 15{\%} training time reduction compared to a standard RL algorithm using auto-regressive generation.

KRLS: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning

emnlp-main.759

2211.16773

Poster

https://aclanthology.org/2023.emnlp-main.760.bib

https://aclanthology.org/2023.emnlp-main.760/

@inproceedings{koto-etal-2023-large, title = "Large Language Models Only Pass Primary School Exams in {I}ndonesia: A Comprehensive Test on {I}ndo{MMLU}", author = "Koto, Fajri and Aisyah, Nurul and Li, Haonan and Baldwin, Timothy", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.760", doi = "10.18653/v1/2023.emnlp-main.760", pages = "12359--12374", abstract = "Although large language models (LLMs) are often pre-trained on large-scale multilingual texts, their reasoning abilities and real-world knowledge are mainly evaluated based on English datasets. Assessing LLM capabilities beyond English is increasingly vital but hindered due to the lack of suitable datasets. In this work, we introduce IndoMMLU, the first multi-task language understanding benchmark for Indonesian culture and languages, which consists of questions from primary school to university entrance exams in Indonesia. By employing professional teachers, we obtain 14,981 questions across 64 tasks and education levels, with 46{\%} of the questions focusing on assessing proficiency in the Indonesian language and knowledge of nine local languages and cultures in Indonesia. Our empirical evaluations show that GPT-3.5 only manages to pass the Indonesian primary school level, with limited knowledge of local Indonesian languages and culture. Other smaller models such as BLOOMZ and Falcon perform at even lower levels.", }

Although large language models (LLMs) are often pre-trained on large-scale multilingual texts, their reasoning abilities and real-world knowledge are mainly evaluated based on English datasets. Assessing LLM capabilities beyond English is increasingly vital but hindered due to the lack of suitable datasets. In this work, we introduce IndoMMLU, the first multi-task language understanding benchmark for Indonesian culture and languages, which consists of questions from primary school to university entrance exams in Indonesia. By employing professional teachers, we obtain 14,981 questions across 64 tasks and education levels, with 46{\%} of the questions focusing on assessing proficiency in the Indonesian language and knowledge of nine local languages and cultures in Indonesia. Our empirical evaluations show that GPT-3.5 only manages to pass the Indonesian primary school level, with limited knowledge of local Indonesian languages and culture. Other smaller models such as BLOOMZ and Falcon perform at even lower levels.

Large Language Models Only Pass Primary School Exams in Indonesia: A Comprehensive Test on IndoMMLU

emnlp-main.760

2310.04928

https://huggingface.co/papers/2310.04928

Oral

https://aclanthology.org/2023.emnlp-main.761.bib

https://aclanthology.org/2023.emnlp-main.761/

@inproceedings{aggarwal-etal-2023-lets, title = "Let{'}s Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with {LLM}s", author = "Aggarwal, Pranjal and Madaan, Aman and Yang, Yiming and {Mausam}", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.761", doi = "10.18653/v1/2023.emnlp-main.761", pages = "12375--12396", abstract = "A popular approach for improving the correctness of output from large language models (LLMs) is Self-Consistency - poll the LLM multiple times and output the most frequent solution. Existing Self-Consistency techniques always generate a constant number of samples per question, where a better approach will be to non-uniformly distribute the available budget based on the amount of agreement in the samples generated so far. In response, we introduce Adaptive-Consistency, a cost-efficient, model-agnostic technique that dynamically adjusts the number of samples per question using a lightweight stopping criterion. Our experiments over 17 reasoning and code generation datasets and three LLMs demonstrate that Adaptive-Consistency reduces sample budget by up to 7.9 times with an average accuracy drop of less than 0.1{\%}", }

A popular approach for improving the correctness of output from large language models (LLMs) is Self-Consistency - poll the LLM multiple times and output the most frequent solution. Existing Self-Consistency techniques always generate a constant number of samples per question, where a better approach will be to non-uniformly distribute the available budget based on the amount of agreement in the samples generated so far. In response, we introduce Adaptive-Consistency, a cost-efficient, model-agnostic technique that dynamically adjusts the number of samples per question using a lightweight stopping criterion. Our experiments over 17 reasoning and code generation datasets and three LLMs demonstrate that Adaptive-Consistency reduces sample budget by up to 7.9 times with an average accuracy drop of less than 0.1{\%}

Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMs

emnlp-main.761

2305.11860

https://huggingface.co/papers/2305.11860

Oral

https://aclanthology.org/2023.emnlp-main.762.bib

https://aclanthology.org/2023.emnlp-main.762/

@inproceedings{cheng-etal-2023-bridging, title = "Bridging Information-Theoretic and Geometric Compression in Language Models", author = "Cheng, Emily and Kervadec, Corentin and Baroni, Marco", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.762", doi = "10.18653/v1/2023.emnlp-main.762", pages = "12397--12420", abstract = "For a language model (LM) to faithfully model human language, it must compress vast, potentially infinite information into relatively few dimensions. We propose analyzing compression in (pre-trained) LMs from two points of view: geometric and information-theoretic. We demonstrate that the two views are highly correlated, such that the intrinsic geometric dimension of linguistic data predicts their coding length under the LM. We then show that, in turn, high compression of a linguistic dataset predicts rapid adaptation to that dataset, confirming that being able to compress linguistic information is an important part of successful LM performance. As a practical byproduct of our analysis, we evaluate a battery of intrinsic dimension estimators for the first time on linguistic data, showing that only some encapsulate the relationship between information-theoretic compression, geometric compression, and ease-of-adaptation.", }

For a language model (LM) to faithfully model human language, it must compress vast, potentially infinite information into relatively few dimensions. We propose analyzing compression in (pre-trained) LMs from two points of view: geometric and information-theoretic. We demonstrate that the two views are highly correlated, such that the intrinsic geometric dimension of linguistic data predicts their coding length under the LM. We then show that, in turn, high compression of a linguistic dataset predicts rapid adaptation to that dataset, confirming that being able to compress linguistic information is an important part of successful LM performance. As a practical byproduct of our analysis, we evaluate a battery of intrinsic dimension estimators for the first time on linguistic data, showing that only some encapsulate the relationship between information-theoretic compression, geometric compression, and ease-of-adaptation.

Bridging Information-Theoretic and Geometric Compression in Language Models

emnlp-main.762

2310.13620

https://huggingface.co/papers/2310.13620

Oral

https://aclanthology.org/2023.emnlp-main.763.bib

https://aclanthology.org/2023.emnlp-main.763/

@inproceedings{yu-etal-2023-pre, title = "Pre-training Language Models for Comparative Reasoning", author = "Yu, Mengxia and Zhang, Zhihan and Yu, Wenhao and Jiang, Meng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.763", doi = "10.18653/v1/2023.emnlp-main.763", pages = "12421--12433", abstract = "Comparative reasoning is a process of comparing objects, concepts, or entities to draw conclusions, which constitutes a fundamental cognitive ability. In this paper, we propose a novel framework to pre-train language models for enhancing their abilities of comparative reasoning over texts. While there have been approaches for NLP tasks that require comparative reasoning, they suffer from costly manual data labeling and limited generalizability to different tasks. Our approach introduces a novel method of collecting scalable data for text-based entity comparison, which leverages both structured and unstructured data. Moreover, we present a framework of pre-training language models via three novel objectives on comparative reasoning. Evaluation on downstream tasks including comparative question answering, question generation, and summarization shows that our pre-training framework significantly improves the comparative reasoning abilities of language models, especially under low-resource conditions. This work also releases the first integrated benchmark for comparative reasoning.", }

Comparative reasoning is a process of comparing objects, concepts, or entities to draw conclusions, which constitutes a fundamental cognitive ability. In this paper, we propose a novel framework to pre-train language models for enhancing their abilities of comparative reasoning over texts. While there have been approaches for NLP tasks that require comparative reasoning, they suffer from costly manual data labeling and limited generalizability to different tasks. Our approach introduces a novel method of collecting scalable data for text-based entity comparison, which leverages both structured and unstructured data. Moreover, we present a framework of pre-training language models via three novel objectives on comparative reasoning. Evaluation on downstream tasks including comparative question answering, question generation, and summarization shows that our pre-training framework significantly improves the comparative reasoning abilities of language models, especially under low-resource conditions. This work also releases the first integrated benchmark for comparative reasoning.

Pre-training Language Models for Comparative Reasoning

emnlp-main.763

2305.14457

https://huggingface.co/papers/2305.14457

Oral

https://aclanthology.org/2023.emnlp-main.764.bib

https://aclanthology.org/2023.emnlp-main.764/

@inproceedings{geng-etal-2023-improved, title = "Improved Pseudo Data for Machine Translation Quality Estimation with Constrained Beam Search", author = "Geng, Xiang and Zhang, Yu and Lai, Zhejian and She, Shuaijie and Zou, Wei and Tao, Shimin and Yang, Hao and Chen, Jiajun and Huang, Shujian", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.764", doi = "10.18653/v1/2023.emnlp-main.764", pages = "12434--12447", abstract = "Machine translation (MT) quality estimation (QE) is a crucial task to estimate the quality of MT outputs when reference translations are unavailable. Many studies focus on generating pseudo data using large parallel corpus and achieve remarkable success in the supervised setting. However, pseudo data solutions are less satisfying in unsupervised scenarios because the pseudo labels are inaccurate or the pseudo translations differ from the real ones. To address these problems, we propose to generate pseudo data using the MT model with constrained beam search (CBSQE). CBSQE preserves the reference parts with high MT probabilities as correct translations, while the rest parts as the wrong ones for MT generation. Therefore, CBSQE can reduce the false negative labels caused by synonyms. Overall, beam search will prefer a more real hypothesis with a higher MT generation likelihood. Extensive experiments demonstrate that CBSQE outperforms strong baselines in both supervised and unsupervised settings. Analyses further show the superiority of CBSQE. The code is available at https://github.com/NJUNLP/njuqe.", }

Machine translation (MT) quality estimation (QE) is a crucial task to estimate the quality of MT outputs when reference translations are unavailable. Many studies focus on generating pseudo data using large parallel corpus and achieve remarkable success in the supervised setting. However, pseudo data solutions are less satisfying in unsupervised scenarios because the pseudo labels are inaccurate or the pseudo translations differ from the real ones. To address these problems, we propose to generate pseudo data using the MT model with constrained beam search (CBSQE). CBSQE preserves the reference parts with high MT probabilities as correct translations, while the rest parts as the wrong ones for MT generation. Therefore, CBSQE can reduce the false negative labels caused by synonyms. Overall, beam search will prefer a more real hypothesis with a higher MT generation likelihood. Extensive experiments demonstrate that CBSQE outperforms strong baselines in both supervised and unsupervised settings. Analyses further show the superiority of CBSQE. The code is available at https://github.com/NJUNLP/njuqe.

Improved Pseudo Data for Machine Translation Quality Estimation with Constrained Beam Search

emnlp-main.764

Poster

https://aclanthology.org/2023.emnlp-main.765.bib

https://aclanthology.org/2023.emnlp-main.765/

@inproceedings{morris-etal-2023-text, title = "Text Embeddings Reveal (Almost) As Much As Text", author = "Morris, John and Kuleshov, Volodymyr and Shmatikov, Vitaly and Rush, Alexander", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.765", doi = "10.18653/v1/2023.emnlp-main.765", pages = "12448--12460", abstract = "How much private information do text embeddings reveal about the original text? We investigate the problem of embedding \textit{inversion}, reconstructing the full text represented in dense text embeddings. We frame the problem as controlled generation: generating text that, when reembedded, is close to a fixed point in latent space. We find that although a naive model conditioned on the embedding performs poorly, a multi-step method that iteratively corrects and re-embeds text is able to recover 92{\%} of 32-token text inputs exactly. We train our model to decode text embeddings from two state-of-the-art embedding models, and also show that our model can recover important personal information (full names) from a dataset of clinical notes.", }

How much private information do text embeddings reveal about the original text? We investigate the problem of embedding \textit{inversion}, reconstructing the full text represented in dense text embeddings. We frame the problem as controlled generation: generating text that, when reembedded, is close to a fixed point in latent space. We find that although a naive model conditioned on the embedding performs poorly, a multi-step method that iteratively corrects and re-embeds text is able to recover 92{\%} of 32-token text inputs exactly. We train our model to decode text embeddings from two state-of-the-art embedding models, and also show that our model can recover important personal information (full names) from a dataset of clinical notes.

Text Embeddings Reveal (Almost) As Much As Text

emnlp-main.765

2310.06816

https://huggingface.co/papers/2310.06816

Poster

https://aclanthology.org/2023.emnlp-main.766.bib

https://aclanthology.org/2023.emnlp-main.766/

@inproceedings{wang-etal-2023-autotrial, title = "{A}uto{T}rial: Prompting Language Models for Clinical Trial Design", author = "Wang, Zifeng and Xiao, Cao and Sun, Jimeng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.766", doi = "10.18653/v1/2023.emnlp-main.766", pages = "12461--12472", abstract = "Clinical trials are critical for drug development. Constructing the appropriate eligibility criteria (i.e., the inclusion/exclusion criteria for patient recruitment) is essential for the trial{'}s success. Proper design of clinical trial protocols should consider similar precedent trials and their eligibility criteria to ensure sufficient patient coverage. In this paper, we present a method named AutoTrial to aid the design of clinical eligibility criteria using language models. It allows (1) controllable generation under instructions via a hybrid of discrete and neural prompting, (2) scalable knowledge incorporation via in-context learning, and (3) explicit reasoning chains to provide rationales for understanding the outputs. Experiments on over 70K clinical trials verify that AutoTrial generates high-quality criteria texts that are fluent and coherent and with high accuracy in capturing the relevant clinical concepts to the target trial. It is noteworthy that our method, with a much smaller parameter size, gains around 60{\%} winning rate against the GPT-3.5 baselines via human evaluations.", }

Clinical trials are critical for drug development. Constructing the appropriate eligibility criteria (i.e., the inclusion/exclusion criteria for patient recruitment) is essential for the trial{'}s success. Proper design of clinical trial protocols should consider similar precedent trials and their eligibility criteria to ensure sufficient patient coverage. In this paper, we present a method named AutoTrial to aid the design of clinical eligibility criteria using language models. It allows (1) controllable generation under instructions via a hybrid of discrete and neural prompting, (2) scalable knowledge incorporation via in-context learning, and (3) explicit reasoning chains to provide rationales for understanding the outputs. Experiments on over 70K clinical trials verify that AutoTrial generates high-quality criteria texts that are fluent and coherent and with high accuracy in capturing the relevant clinical concepts to the target trial. It is noteworthy that our method, with a much smaller parameter size, gains around 60{\%} winning rate against the GPT-3.5 baselines via human evaluations.

AutoTrial: Prompting Language Models for Clinical Trial Design

emnlp-main.766

2305.11366

Poster

https://aclanthology.org/2023.emnlp-main.767.bib

https://aclanthology.org/2023.emnlp-main.767/

@inproceedings{cheng-vlachos-2023-faster, title = "Faster Minimum {B}ayes Risk Decoding with Confidence-based Pruning", author = "Cheng, Julius and Vlachos, Andreas", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.767", doi = "10.18653/v1/2023.emnlp-main.767", pages = "12473--12480", abstract = "Minimum Bayes risk (MBR) decoding outputs the hypothesis with the highest expected utility over the model distribution for some utility function. It has been shown to improve accuracy over beam search in conditional language generation problems and especially neural machine translation, in both human and automatic evaluations. However, the standard sampling-based algorithm for MBR is substantially more computationally expensive than beam search, requiring a large number of samples as well as a quadratic number of calls to the utility function, limiting its applicability. We describe an algorithm for MBR which gradually grows the number of samples used to estimate the utility while pruning hypotheses that are unlikely to have the highest utility according to confidence estimates obtained with bootstrap sampling. Our method requires fewer samples and drastically reduces the number of calls to the utility function compared to standard MBR while being statistically indistinguishable in terms of accuracy. We demonstrate the effectiveness of our approach in experiments on three language pairs, using chrF++ and COMET as utility/evaluation metrics.", }

Minimum Bayes risk (MBR) decoding outputs the hypothesis with the highest expected utility over the model distribution for some utility function. It has been shown to improve accuracy over beam search in conditional language generation problems and especially neural machine translation, in both human and automatic evaluations. However, the standard sampling-based algorithm for MBR is substantially more computationally expensive than beam search, requiring a large number of samples as well as a quadratic number of calls to the utility function, limiting its applicability. We describe an algorithm for MBR which gradually grows the number of samples used to estimate the utility while pruning hypotheses that are unlikely to have the highest utility according to confidence estimates obtained with bootstrap sampling. Our method requires fewer samples and drastically reduces the number of calls to the utility function compared to standard MBR while being statistically indistinguishable in terms of accuracy. We demonstrate the effectiveness of our approach in experiments on three language pairs, using chrF++ and COMET as utility/evaluation metrics.

Faster Minimum Bayes Risk Decoding with Confidence-based Pruning

emnlp-main.767

2311.14919

Poster

https://aclanthology.org/2023.emnlp-main.768.bib

https://aclanthology.org/2023.emnlp-main.768/

@inproceedings{zhou-etal-2023-enhancing-generative, title = "Enhancing Generative Retrieval with Reinforcement Learning from Relevance Feedback", author = "Zhou, Yujia and Dou, Zhicheng and Wen, Ji-Rong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.768", doi = "10.18653/v1/2023.emnlp-main.768", pages = "12481--12490", abstract = "The recent advent of end-to-end generative retrieval marks a significant shift in document retrieval methods, leveraging differentiable search indexes to directly produce relevant document identifiers (docids) in response to a specific query. Nevertheless, this approach faces two fundamental challenges: (i) a discrepancy between the token-level probabilistic optimization and the broader document-level relevance estimation; (ii) an overemphasis on top-1 results at the expense of overall ranking quality. To tackle these challenges, we propose a generative retrieval model with reinforcement learning from relevance feedback, which aims to align token-level docid generation with document-level relevance estimation. The training process incorporates three stages: supervised fine-tuning, relevance reward model training, and reinforced learning-to-rank from relevance feedback. To train a high-quality reward model, we define {``}relevance{''} under three progressive scenarios, which collectively offer a comprehensive evaluation of the document relevance. Experiments conducted on two benchmark datasets demonstrate the effectiveness of our proposed approach.", }

The recent advent of end-to-end generative retrieval marks a significant shift in document retrieval methods, leveraging differentiable search indexes to directly produce relevant document identifiers (docids) in response to a specific query. Nevertheless, this approach faces two fundamental challenges: (i) a discrepancy between the token-level probabilistic optimization and the broader document-level relevance estimation; (ii) an overemphasis on top-1 results at the expense of overall ranking quality. To tackle these challenges, we propose a generative retrieval model with reinforcement learning from relevance feedback, which aims to align token-level docid generation with document-level relevance estimation. The training process incorporates three stages: supervised fine-tuning, relevance reward model training, and reinforced learning-to-rank from relevance feedback. To train a high-quality reward model, we define {``}relevance{''} under three progressive scenarios, which collectively offer a comprehensive evaluation of the document relevance. Experiments conducted on two benchmark datasets demonstrate the effectiveness of our proposed approach.

Enhancing Generative Retrieval with Reinforcement Learning from Relevance Feedback

emnlp-main.768

Poster

https://aclanthology.org/2023.emnlp-main.769.bib

https://aclanthology.org/2023.emnlp-main.769/

@inproceedings{li-etal-2023-multi-source-probing, title = "Multi-Source Probing for Open-Domain Conversational Understanding", author = "Li, Yuanxi and Zhou, Hao and Zhou, Jie and Huang, Minlie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.769", doi = "10.18653/v1/2023.emnlp-main.769", pages = "12491--12505", abstract = "Dialogue comprehension and generation are vital to the success of open-domain dialogue systems. Although pre-trained generative conversation models have made significant progress in generating fluent responses, people have difficulty judging whether they understand and efficiently model the contextual information of the conversation. In this study, we propose a Multi-Source Probing (MSP) method to probe the dialogue comprehension abilities of open-domain dialogue models. MSP aggregates features from multiple sources to accomplish diverse task goals and conducts downstream tasks in a generative manner that is consistent with dialogue model pre-training to leverage model capabilities. We conduct probing experiments on seven tasks that require various dialogue comprehension skills, based on the internal representations encoded by dialogue models. Experimental results show that open-domain dialogue models can encode semantic information in the intermediate hidden states, which facilitates dialogue comprehension tasks. Models of different scales and structures possess different conversational understanding capabilities. Our findings encourage a comprehensive evaluation and design of open-domain dialogue models.", }

Dialogue comprehension and generation are vital to the success of open-domain dialogue systems. Although pre-trained generative conversation models have made significant progress in generating fluent responses, people have difficulty judging whether they understand and efficiently model the contextual information of the conversation. In this study, we propose a Multi-Source Probing (MSP) method to probe the dialogue comprehension abilities of open-domain dialogue models. MSP aggregates features from multiple sources to accomplish diverse task goals and conducts downstream tasks in a generative manner that is consistent with dialogue model pre-training to leverage model capabilities. We conduct probing experiments on seven tasks that require various dialogue comprehension skills, based on the internal representations encoded by dialogue models. Experimental results show that open-domain dialogue models can encode semantic information in the intermediate hidden states, which facilitates dialogue comprehension tasks. Models of different scales and structures possess different conversational understanding capabilities. Our findings encourage a comprehensive evaluation and design of open-domain dialogue models.

Multi-Source Probing for Open-Domain Conversational Understanding

emnlp-main.769

Poster

https://aclanthology.org/2023.emnlp-main.770.bib

https://aclanthology.org/2023.emnlp-main.770/

@inproceedings{shi-etal-2023-hallucination, title = "Hallucination Mitigation in Natural Language Generation from Large-Scale Open-Domain Knowledge Graphs", author = "Shi, Xiao and Zhu, Zhengyuan and Zhang, Zeyu and Li, Chengkai", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.770", doi = "10.18653/v1/2023.emnlp-main.770", pages = "12506--12521", abstract = "In generating natural language descriptions for knowledge graph triples, prior works used either small-scale, human-annotated datasets or datasets with limited variety of graph shapes, e.g., those having mostly star graphs. Graph-to-text models trained and evaluated on such datasets are largely not assessed for more realistic large-scale, open-domain settings. We introduce a new dataset, GraphNarrative, to fill this gap. Fine-tuning transformer-based pre-trained language models has achieved state-of-the-art performance among graph-to-text models. However, this method suffers from information hallucination{---}the generated text may contain fabricated facts not present in input graphs. We propose a novel approach that, given a graph-sentence pair in GraphNarrative, trims the sentence to eliminate portions that are not present in the corresponding graph, by utilizing the sentence{'}s dependency parse tree. Our experiment results verify this approach using models trained on GraphNarrative and existing datasets. The dataset, source code, and trained models are released at https://github.com/idirlab/graphnarrator.", }

In generating natural language descriptions for knowledge graph triples, prior works used either small-scale, human-annotated datasets or datasets with limited variety of graph shapes, e.g., those having mostly star graphs. Graph-to-text models trained and evaluated on such datasets are largely not assessed for more realistic large-scale, open-domain settings. We introduce a new dataset, GraphNarrative, to fill this gap. Fine-tuning transformer-based pre-trained language models has achieved state-of-the-art performance among graph-to-text models. However, this method suffers from information hallucination{---}the generated text may contain fabricated facts not present in input graphs. We propose a novel approach that, given a graph-sentence pair in GraphNarrative, trims the sentence to eliminate portions that are not present in the corresponding graph, by utilizing the sentence{'}s dependency parse tree. Our experiment results verify this approach using models trained on GraphNarrative and existing datasets. The dataset, source code, and trained models are released at https://github.com/idirlab/graphnarrator.

Hallucination Mitigation in Natural Language Generation from Large-Scale Open-Domain Knowledge Graphs

emnlp-main.770

Poster

https://aclanthology.org/2023.emnlp-main.771.bib

https://aclanthology.org/2023.emnlp-main.771/

@inproceedings{ni-etal-2023-multi, title = "Multi-Source Multi-Type Knowledge Exploration and Exploitation for Dialogue Generation", author = "Ni, Xuanfan and Dai, Hongliang and Ren, Zhaochun and Li, Piji", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.771", pages = "12522--12537", abstract = "Open-domain multi-turn dialogue generation encounters the significant challenge of lacking various types of knowledge from diverse sources. Existing models typically focus on identifying specific types of dialogue knowledge and utilize corresponding datasets for training. However, this approach often leads to limited generalization capabilities and increased computational resource requirements. Recently, large language models (LLMs) have shown impressive performance on natural language processing tasks. To harness the knowledge storage of LLMs, we propose a framework named KnowEE that explores multi-source multi-type knowledge from LLMs by leveraging diverse datasets and then exploits the obtained knowledge for response generation. Our framework comprises two phases: First, we leverage five external datasets encompassing various types of knowledge to extract the most relevant samples to the dialogue context which are served as prompts to generate corresponding type of knowledge; Second, we inject the acquired knowledge into the ongoing dialogue context in fine-grained and coarse-grained manners, which is then fed into LLMs to generate the final dialogue response. Both automatic and manual evaluation results validate the effectiveness of our framework in exploring and exploiting multi-source multi-type knowledge to generate coherent, informative, and fluent responses.", }

Open-domain multi-turn dialogue generation encounters the significant challenge of lacking various types of knowledge from diverse sources. Existing models typically focus on identifying specific types of dialogue knowledge and utilize corresponding datasets for training. However, this approach often leads to limited generalization capabilities and increased computational resource requirements. Recently, large language models (LLMs) have shown impressive performance on natural language processing tasks. To harness the knowledge storage of LLMs, we propose a framework named KnowEE that explores multi-source multi-type knowledge from LLMs by leveraging diverse datasets and then exploits the obtained knowledge for response generation. Our framework comprises two phases: First, we leverage five external datasets encompassing various types of knowledge to extract the most relevant samples to the dialogue context which are served as prompts to generate corresponding type of knowledge; Second, we inject the acquired knowledge into the ongoing dialogue context in fine-grained and coarse-grained manners, which is then fed into LLMs to generate the final dialogue response. Both automatic and manual evaluation results validate the effectiveness of our framework in exploring and exploiting multi-source multi-type knowledge to generate coherent, informative, and fluent responses.

Multi-Source Multi-Type Knowledge Exploration and Exploitation for Dialogue Generation

emnlp-main.771

Poster

https://aclanthology.org/2023.emnlp-main.772.bib

https://aclanthology.org/2023.emnlp-main.772/

@inproceedings{lutati-etal-2023-focus, title = "Focus Your Attention (with Adaptive {IIR} Filters)", author = "Lutati, Shahar and Zimerman, Itamar and Wolf, Lior", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.772", doi = "10.18653/v1/2023.emnlp-main.772", pages = "12538--12549", abstract = "We present a new layer in which dynamic (i.e., input-dependent) Infinite Impulse Response (IIR) filters of order two are used to process the input sequence prior to applying conventional attention. The input is split into chunks, and the coefficients of these filters are determined based on previous chunks to maintain causality. Despite their relatively low order, the causal adaptive filters are shown to focus attention on the relevant sequence elements. The new layer is grounded in control theory, and is shown to generalize diagonal state-space layers. The layer performs on-par with state-of-the-art networks, with a fraction of their parameters and with time complexity that is sub-quadratic with input size. The obtained layer is favorable to layers such as Heyna, GPT2, and Mega, both with respect to the number of parameters and the obtained level of performance on multiple long-range sequence problems.", }

We present a new layer in which dynamic (i.e., input-dependent) Infinite Impulse Response (IIR) filters of order two are used to process the input sequence prior to applying conventional attention. The input is split into chunks, and the coefficients of these filters are determined based on previous chunks to maintain causality. Despite their relatively low order, the causal adaptive filters are shown to focus attention on the relevant sequence elements. The new layer is grounded in control theory, and is shown to generalize diagonal state-space layers. The layer performs on-par with state-of-the-art networks, with a fraction of their parameters and with time complexity that is sub-quadratic with input size. The obtained layer is favorable to layers such as Heyna, GPT2, and Mega, both with respect to the number of parameters and the obtained level of performance on multiple long-range sequence problems.

Focus Your Attention (with Adaptive IIR Filters)

emnlp-main.772

2305.14952

Oral

https://aclanthology.org/2023.emnlp-main.773.bib

https://aclanthology.org/2023.emnlp-main.773/

@inproceedings{maekawa-imai-2023-identifying, title = "Identifying Statements Crucial for Awareness of Interpretive Nonsense to Prevent Communication Breakdowns", author = "Maekawa, Tomoyuki and Imai, Michita", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.773", doi = "10.18653/v1/2023.emnlp-main.773", pages = "12550--12566", abstract = "During remote conversations, communication breakdowns often occur when a listener misses certain statements. Our objective is to prevent such breakdowns by identifying Statements Crucial for Awareness of Interpretive Nonsense (SCAINs). If a listener misses a SCAIN, s/he may interpret subsequent statements differently from the speaker{'}s intended meaning. To identify SCAINs, we adopt a unique approach where we create a dialogue by omitting two consecutive statements from the original dialogue and then generate text to make the following statement more specific. The novelty of the proposed method lies in simulating missing information by processing text with omissions. We validate the effectiveness of SCAINs through evaluation using a dialogue dataset. Furthermore, we demonstrate that SCAINs cannot be identified as merely important statements, highlighting the uniqueness of our proposed method.", }

During remote conversations, communication breakdowns often occur when a listener misses certain statements. Our objective is to prevent such breakdowns by identifying Statements Crucial for Awareness of Interpretive Nonsense (SCAINs). If a listener misses a SCAIN, s/he may interpret subsequent statements differently from the speaker{'}s intended meaning. To identify SCAINs, we adopt a unique approach where we create a dialogue by omitting two consecutive statements from the original dialogue and then generate text to make the following statement more specific. The novelty of the proposed method lies in simulating missing information by processing text with omissions. We validate the effectiveness of SCAINs through evaluation using a dialogue dataset. Furthermore, we demonstrate that SCAINs cannot be identified as merely important statements, highlighting the uniqueness of our proposed method.

Identifying Statements Crucial for Awareness of Interpretive Nonsense to Prevent Communication Breakdowns

emnlp-main.773

Poster

https://aclanthology.org/2023.emnlp-main.774.bib

https://aclanthology.org/2023.emnlp-main.774/

@inproceedings{zhang-etal-2023-multilingual, title = "Multilingual Large Language Models Are Not (Yet) Code-Switchers", author = "Zhang, Ruochen and Cahyawijaya, Samuel and Cruz, Jan Christian Blaise and Winata, Genta and Aji, Alham Fikri", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.774", doi = "10.18653/v1/2023.emnlp-main.774", pages = "12567--12582", abstract = "Multilingual Large Language Models (LLMs) have recently shown great capabilities in a wide range of tasks, exhibiting state-of-the-art performance through zero-shot or few-shot prompting methods. While there have been extensive studies on their abilities in monolingual tasks, the investigation of their potential in the context of code-switching (CSW), the practice of alternating languages within an utterance, remains relatively uncharted. In this paper, we provide a comprehensive empirical analysis of various multilingual LLMs, benchmarking their performance across four tasks: sentiment analysis, machine translation, summarization and word-level language identification. Our results indicate that despite multilingual LLMs exhibiting promising outcomes in certain tasks using zero or few-shot prompting, they still underperform in comparison to fine-tuned models of much smaller scales. We argue that current {``}multilingualism{'} in LLMs does not inherently imply proficiency with code-switching texts, calling for future research to bridge this discrepancy.", }

Multilingual Large Language Models (LLMs) have recently shown great capabilities in a wide range of tasks, exhibiting state-of-the-art performance through zero-shot or few-shot prompting methods. While there have been extensive studies on their abilities in monolingual tasks, the investigation of their potential in the context of code-switching (CSW), the practice of alternating languages within an utterance, remains relatively uncharted. In this paper, we provide a comprehensive empirical analysis of various multilingual LLMs, benchmarking their performance across four tasks: sentiment analysis, machine translation, summarization and word-level language identification. Our results indicate that despite multilingual LLMs exhibiting promising outcomes in certain tasks using zero or few-shot prompting, they still underperform in comparison to fine-tuned models of much smaller scales. We argue that current {``}multilingualism{'} in LLMs does not inherently imply proficiency with code-switching texts, calling for future research to bridge this discrepancy.

Multilingual Large Language Models Are Not (Yet) Code-Switchers

emnlp-main.774

2305.14235

https://huggingface.co/papers/2305.14235

Oral

https://aclanthology.org/2023.emnlp-main.775.bib

https://aclanthology.org/2023.emnlp-main.775/

@inproceedings{dao-etal-2023-reinforced, title = "Reinforced Target-driven Conversational Promotion", author = "Dao, Huy and Liao, Lizi and Le, Dung and Nie, Yuxiang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.775", doi = "10.18653/v1/2023.emnlp-main.775", pages = "12583--12596", abstract = "The ability to proactively engage with users towards pitching products is highly desired for conversational assistants. However, existing conversational recommendation methods overemphasize on acquiring user preferences while ignore the strategic planning for nudging users towards accepting a designated item. Hence, these methods fail to promote specified items with engaging responses. In this work, we propose a Reinforced Target-driven Conversational Promotion (RTCP) framework for conversational promotion. RTCP integrates short-term and long-term planning via a balanced gating mechanism. Inside which, the dialogue actions are predicted via a knowledge-integrated multi-head attention and guided via reinforcement learning rewards. RTCP then employs action-guided prefix tuning to generate relevant responses. Experimental results demonstrate that our model outperforms state-of-the-art models on both automatic metrics and human evaluation. Moreover, RTCP has a strong capability in quickly adapting to unseen scenarios just by updating prefix parameters without re-training the whole model.", }

The ability to proactively engage with users towards pitching products is highly desired for conversational assistants. However, existing conversational recommendation methods overemphasize on acquiring user preferences while ignore the strategic planning for nudging users towards accepting a designated item. Hence, these methods fail to promote specified items with engaging responses. In this work, we propose a Reinforced Target-driven Conversational Promotion (RTCP) framework for conversational promotion. RTCP integrates short-term and long-term planning via a balanced gating mechanism. Inside which, the dialogue actions are predicted via a knowledge-integrated multi-head attention and guided via reinforcement learning rewards. RTCP then employs action-guided prefix tuning to generate relevant responses. Experimental results demonstrate that our model outperforms state-of-the-art models on both automatic metrics and human evaluation. Moreover, RTCP has a strong capability in quickly adapting to unseen scenarios just by updating prefix parameters without re-training the whole model.

Reinforced Target-driven Conversational Promotion

emnlp-main.775

Poster

https://aclanthology.org/2023.emnlp-main.776.bib

https://aclanthology.org/2023.emnlp-main.776/

@inproceedings{weinzierl-harabagiu-2023-identification, title = "Identification of Multimodal Stance Towards Frames of Communication", author = "Weinzierl, Maxwell and Harabagiu, Sanda", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.776", doi = "10.18653/v1/2023.emnlp-main.776", pages = "12597--12609", abstract = "Frames of communication are often evoked in multimedia documents. When an author decides to add an image to a text, one or both of the modalities may evoke a communication frame. Moreover, when evoking the frame, the author also conveys her/his stance towards the frame. Until now, determining if the author is in favor of, against or has no stance towards the frame was performed automatically only when processing texts. This is due to the absence of stance annotations on multimedia documents. In this paper we introduce MMVax-Stance, a dataset of 11,300 multimedia documents retrieved from social media, which have stance annotations towards 113 different frames of communication. This dataset allowed us to experiment with several models of multimedia stance detection, which revealed important interactions between texts and images in the inference of stance towards communication frames. When inferring the text/image relations, a set of 46,606 synthetic examples of multimodal documents with known stance was generated. This greatly impacted the quality of identifying multimedia stance, yielding an improvement of 20{\%} in F1-score.", }

Frames of communication are often evoked in multimedia documents. When an author decides to add an image to a text, one or both of the modalities may evoke a communication frame. Moreover, when evoking the frame, the author also conveys her/his stance towards the frame. Until now, determining if the author is in favor of, against or has no stance towards the frame was performed automatically only when processing texts. This is due to the absence of stance annotations on multimedia documents. In this paper we introduce MMVax-Stance, a dataset of 11,300 multimedia documents retrieved from social media, which have stance annotations towards 113 different frames of communication. This dataset allowed us to experiment with several models of multimedia stance detection, which revealed important interactions between texts and images in the inference of stance towards communication frames. When inferring the text/image relations, a set of 46,606 synthetic examples of multimodal documents with known stance was generated. This greatly impacted the quality of identifying multimedia stance, yielding an improvement of 20{\%} in F1-score.

Identification of Multimodal Stance Towards Frames of Communication

emnlp-main.776

Oral

https://aclanthology.org/2023.emnlp-main.777.bib

https://aclanthology.org/2023.emnlp-main.777/

@inproceedings{zhang-etal-2023-unsupervised, title = "Unsupervised Sounding Pixel Learning", author = "Zhang, Yining and Ji, Yanli and Yang, Yang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.777", doi = "10.18653/v1/2023.emnlp-main.777", pages = "12610--12620", abstract = "Sounding source localization is a challenging cross-modal task due to the difficulty of cross-modal alignment. Although supervised cross-modal methods achieve encouraging performance, heavy manual annotations are expensive and inefficient. Thus it is valuable and meaningful to develop unsupervised solutions. In this paper, we propose an **U**nsupervised **S**ounding **P**ixel **L**earning (USPL) approach which enables a pixel-level sounding source localization in unsupervised paradigm. We first design a mask augmentation based multi-instance contrastive learning to realize unsupervised cross-modal coarse localization, which aligns audio-visual features to obtain coarse sounding maps. Secondly, we present an *Unsupervised Sounding Map Refinement (SMR)* module which employs the visual semantic affinity learning to explore inter-pixel relations of adjacent coordinate features. It contributes to recovering the boundary of coarse sounding maps and obtaining fine sounding maps. Finally, a *Sounding Pixel Segmentation (SPS)* module is presented to realize audio-supervised semantic segmentation. Extensive experiments are performed on the AVSBench-S4 and VGGSound datasets, exhibiting encouraging results compared with previous SOTA methods.", }

Sounding source localization is a challenging cross-modal task due to the difficulty of cross-modal alignment. Although supervised cross-modal methods achieve encouraging performance, heavy manual annotations are expensive and inefficient. Thus it is valuable and meaningful to develop unsupervised solutions. In this paper, we propose an **U**nsupervised **S**ounding **P**ixel **L**earning (USPL) approach which enables a pixel-level sounding source localization in unsupervised paradigm. We first design a mask augmentation based multi-instance contrastive learning to realize unsupervised cross-modal coarse localization, which aligns audio-visual features to obtain coarse sounding maps. Secondly, we present an *Unsupervised Sounding Map Refinement (SMR)* module which employs the visual semantic affinity learning to explore inter-pixel relations of adjacent coordinate features. It contributes to recovering the boundary of coarse sounding maps and obtaining fine sounding maps. Finally, a *Sounding Pixel Segmentation (SPS)* module is presented to realize audio-supervised semantic segmentation. Extensive experiments are performed on the AVSBench-S4 and VGGSound datasets, exhibiting encouraging results compared with previous SOTA methods.

Unsupervised Sounding Pixel Learning

emnlp-main.777

Oral

https://aclanthology.org/2023.emnlp-main.778.bib

https://aclanthology.org/2023.emnlp-main.778/

@inproceedings{cohen-etal-2023-lm, title = "{LM} vs {LM}: Detecting Factual Errors via Cross Examination", author = "Cohen, Roi and Hamri, May and Geva, Mor and Globerson, Amir", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.778", doi = "10.18653/v1/2023.emnlp-main.778", pages = "12621--12640", abstract = "A prominent weakness of modern language models (LMs) is their tendency to generate factually incorrect text, which hinders their usability. A natural question is whether such factual errors can be detected automatically. Inspired by truth-seeking mechanisms in law, we propose a factuality evaluation framework for LMs that is based on cross-examination. Our key idea is that an incorrect claim is likely to result in inconsistency with other claims that the model generates. To discover such inconsistencies, we facilitate a multi-turn interaction between the LM that generated the claim and another LM (acting as an examiner) which introduces questions to discover inconsistencies. We empirically evaluate our method on factual claims made by multiple recent LMs on four benchmarks, finding that it outperforms existing methods and baselines, often by a large gap. Our results demonstrate the potential of using interacting LMs for capturing factual errors.", }

A prominent weakness of modern language models (LMs) is their tendency to generate factually incorrect text, which hinders their usability. A natural question is whether such factual errors can be detected automatically. Inspired by truth-seeking mechanisms in law, we propose a factuality evaluation framework for LMs that is based on cross-examination. Our key idea is that an incorrect claim is likely to result in inconsistency with other claims that the model generates. To discover such inconsistencies, we facilitate a multi-turn interaction between the LM that generated the claim and another LM (acting as an examiner) which introduces questions to discover inconsistencies. We empirically evaluate our method on factual claims made by multiple recent LMs on four benchmarks, finding that it outperforms existing methods and baselines, often by a large gap. Our results demonstrate the potential of using interacting LMs for capturing factual errors.

LM vs LM: Detecting Factual Errors via Cross Examination

emnlp-main.778

2305.13281

https://huggingface.co/papers/2305.13281

Poster

https://aclanthology.org/2023.emnlp-main.779.bib

https://aclanthology.org/2023.emnlp-main.779/

@inproceedings{van-dijk-etal-2023-large, title = "Large Language Models: The Need for Nuance in Current Debates and a Pragmatic Perspective on Understanding", author = "van Dijk, Bram and Kouwenhoven, Tom and Spruit, Marco and van Duijn, Max Johannes", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.779", doi = "10.18653/v1/2023.emnlp-main.779", pages = "12641--12654", abstract = "Current Large Language Models (LLMs) are unparalleled in their ability to generate grammatically correct, fluent text. LLMs are appearing rapidly, and debates on LLM capacities have taken off, but reflection is lagging behind. Thus, in this position paper, we first zoom in on the debate and critically assess three points recurring in critiques of LLM capacities: i) that LLMs only parrot statistical patterns in the training data; ii) that LLMs master formal but not functional language competence; and iii) that language learning in LLMs cannot inform human language learning. Drawing on empirical and theoretical arguments, we show that these points need more nuance. Second, we outline a pragmatic perspective on the issue of {`}real{'} understanding and intentionality in LLMs. Understanding and intentionality pertain to unobservable mental states we attribute to other humans because they have pragmatic value: they allow us to abstract away from complex underlying mechanics and predict behaviour effectively. We reflect on the circumstances under which it would make sense for humans to similarly attribute mental states to LLMs, thereby outlining a pragmatic philosophical context for LLMs as an increasingly prominent technology in society.", }

Current Large Language Models (LLMs) are unparalleled in their ability to generate grammatically correct, fluent text. LLMs are appearing rapidly, and debates on LLM capacities have taken off, but reflection is lagging behind. Thus, in this position paper, we first zoom in on the debate and critically assess three points recurring in critiques of LLM capacities: i) that LLMs only parrot statistical patterns in the training data; ii) that LLMs master formal but not functional language competence; and iii) that language learning in LLMs cannot inform human language learning. Drawing on empirical and theoretical arguments, we show that these points need more nuance. Second, we outline a pragmatic perspective on the issue of {`}real{'} understanding and intentionality in LLMs. Understanding and intentionality pertain to unobservable mental states we attribute to other humans because they have pragmatic value: they allow us to abstract away from complex underlying mechanics and predict behaviour effectively. We reflect on the circumstances under which it would make sense for humans to similarly attribute mental states to LLMs, thereby outlining a pragmatic philosophical context for LLMs as an increasingly prominent technology in society.

Large Language Models: The Need for Nuance in Current Debates and a Pragmatic Perspective on Understanding

emnlp-main.779

2310.19671

https://huggingface.co/papers/2310.19671

Oral

https://aclanthology.org/2023.emnlp-main.780.bib

https://aclanthology.org/2023.emnlp-main.780/

@inproceedings{zhang-etal-2023-pieclass, title = "{PIEC}lass: Weakly-Supervised Text Classification with Prompting and Noise-Robust Iterative Ensemble Training", author = "Zhang, Yunyi and Jiang, Minhao and Meng, Yu and Zhang, Yu and Han, Jiawei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.780", doi = "10.18653/v1/2023.emnlp-main.780", pages = "12655--12670", abstract = "Weakly-supervised text classification trains a classifier using the label name of each target class as the only supervision, which largely reduces human annotation efforts. Most existing methods first use the label names as static keyword-based features to generate pseudo labels, which are then used for final classifier training. While reasonable, such a commonly adopted framework suffers from two limitations: (1) keywords can have different meanings in different contexts and some text may not have any keyword, so keyword matching can induce noisy and inadequate pseudo labels; (2) the errors made in the pseudo label generation stage will directly propagate to the classifier training stage without a chance of being corrected. In this paper, we propose a new method, PIEClass, consisting of two modules: (1) a pseudo label acquisition module that uses zero-shot prompting of pre-trained language models (PLM) to get pseudo labels based on contextualized text understanding beyond static keyword matching, and (2) a noise-robust iterative ensemble training module that iteratively trains classifiers and updates pseudo labels by utilizing two PLM fine-tuning methods that regularize each other. Extensive experiments show that PIEClass achieves overall better performance than existing strong baselines on seven benchmark datasets and even achieves similar performance to fully-supervised classifiers on sentiment classification tasks.", }

Weakly-supervised text classification trains a classifier using the label name of each target class as the only supervision, which largely reduces human annotation efforts. Most existing methods first use the label names as static keyword-based features to generate pseudo labels, which are then used for final classifier training. While reasonable, such a commonly adopted framework suffers from two limitations: (1) keywords can have different meanings in different contexts and some text may not have any keyword, so keyword matching can induce noisy and inadequate pseudo labels; (2) the errors made in the pseudo label generation stage will directly propagate to the classifier training stage without a chance of being corrected. In this paper, we propose a new method, PIEClass, consisting of two modules: (1) a pseudo label acquisition module that uses zero-shot prompting of pre-trained language models (PLM) to get pseudo labels based on contextualized text understanding beyond static keyword matching, and (2) a noise-robust iterative ensemble training module that iteratively trains classifiers and updates pseudo labels by utilizing two PLM fine-tuning methods that regularize each other. Extensive experiments show that PIEClass achieves overall better performance than existing strong baselines on seven benchmark datasets and even achieves similar performance to fully-supervised classifiers on sentiment classification tasks.

PIEClass: Weakly-Supervised Text Classification with Prompting and Noise-Robust Iterative Ensemble Training

emnlp-main.780

2305.13723

Poster

https://aclanthology.org/2023.emnlp-main.781.bib

https://aclanthology.org/2023.emnlp-main.781/

@inproceedings{dai-etal-2023-meaeq, title = "{M}eae{Q}: Mount Model Extraction Attacks with Efficient Queries", author = "Dai, Chengwei and Lv, Minxuan and Li, Kun and Zhou, Wei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.781", doi = "10.18653/v1/2023.emnlp-main.781", pages = "12671--12684", abstract = "We study model extraction attacks in natural language processing (NLP) where attackers aim to steal victim models by repeatedly querying the open Application Programming Interfaces (APIs). Recent works focus on limited-query budget settings and adopt random sampling or active learning-based sampling strategies on publicly available, unannotated data sources. However, these methods often result in selected queries that lack task relevance and data diversity, leading to limited success in achieving satisfactory results with low query costs. In this paper, we propose MeaeQ (Model extraction attack with efficient Queries), a straightforward yet effective method to address these issues. Specifically, we initially utilize a zero-shot sequence inference classifier, combined with API service information, to filter task-relevant data from a public text corpus instead of a problem domain-specific dataset. Furthermore, we employ a clustering-based data reduction technique to obtain representative data as queries for the attack. Extensive experiments conducted on four benchmark datasets demonstrate that MeaeQ achieves higher functional similarity to the victim model than baselines while requiring fewer queries.", }

We study model extraction attacks in natural language processing (NLP) where attackers aim to steal victim models by repeatedly querying the open Application Programming Interfaces (APIs). Recent works focus on limited-query budget settings and adopt random sampling or active learning-based sampling strategies on publicly available, unannotated data sources. However, these methods often result in selected queries that lack task relevance and data diversity, leading to limited success in achieving satisfactory results with low query costs. In this paper, we propose MeaeQ (Model extraction attack with efficient Queries), a straightforward yet effective method to address these issues. Specifically, we initially utilize a zero-shot sequence inference classifier, combined with API service information, to filter task-relevant data from a public text corpus instead of a problem domain-specific dataset. Furthermore, we employ a clustering-based data reduction technique to obtain representative data as queries for the attack. Extensive experiments conducted on four benchmark datasets demonstrate that MeaeQ achieves higher functional similarity to the victim model than baselines while requiring fewer queries.

MeaeQ: Mount Model Extraction Attacks with Efficient Queries

emnlp-main.781

2310.14047

Poster

https://aclanthology.org/2023.emnlp-main.782.bib

https://aclanthology.org/2023.emnlp-main.782/

@inproceedings{kim-etal-2023-cot, title = "The {C}o{T} Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuning", author = "Kim, Seungone and Joo, Se and Kim, Doyoung and Jang, Joel and Ye, Seonghyeon and Shin, Jamin and Seo, Minjoon", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.782", doi = "10.18653/v1/2023.emnlp-main.782", pages = "12685--12708", abstract = "Language models (LMs) with less than 100B parameters are known to perform poorly on chain-of-thought (CoT) reasoning in contrast to large LMs when solving unseen tasks. In this work, we aim to equip smaller LMs with the step-by-step reasoning capability by instruction tuning with CoT rationales. In order to achieve this goal, we first introduce a new instruction-tuning dataset called the CoT Collection, which augments the existing Flan Collection (including only 9 CoT tasks) with additional 1.84 million rationales across 1,060 tasks. We show that CoT fine-tuning Flan-T5 (3B {\&} 11B) with CoT Collection enables smaller LMs to have better CoT capabilities on unseen tasks. On the BIG-Bench-Hard (BBH) benchmark, we report an average improvement of +4.34{\%} (Flan-T5 3B) and +2.60{\%} (Flan-T5 11B), in terms of zero-shot task accuracy. Furthermore, we show that instruction tuning with CoT Collection allows LMs to possess stronger few-shot learning capabilities on 4 domain-specific tasks, resulting in an improvement of +2.24{\%} (Flan-T5 3B) and +2.37{\%} (Flan-T5 11B), even outperforming ChatGPT utilizing demonstrations until the max length by a +13.98{\%} margin. Our code, the CoT Collection data, and model checkpoints are publicly available.", }

Language models (LMs) with less than 100B parameters are known to perform poorly on chain-of-thought (CoT) reasoning in contrast to large LMs when solving unseen tasks. In this work, we aim to equip smaller LMs with the step-by-step reasoning capability by instruction tuning with CoT rationales. In order to achieve this goal, we first introduce a new instruction-tuning dataset called the CoT Collection, which augments the existing Flan Collection (including only 9 CoT tasks) with additional 1.84 million rationales across 1,060 tasks. We show that CoT fine-tuning Flan-T5 (3B {\&} 11B) with CoT Collection enables smaller LMs to have better CoT capabilities on unseen tasks. On the BIG-Bench-Hard (BBH) benchmark, we report an average improvement of +4.34{\%} (Flan-T5 3B) and +2.60{\%} (Flan-T5 11B), in terms of zero-shot task accuracy. Furthermore, we show that instruction tuning with CoT Collection allows LMs to possess stronger few-shot learning capabilities on 4 domain-specific tasks, resulting in an improvement of +2.24{\%} (Flan-T5 3B) and +2.37{\%} (Flan-T5 11B), even outperforming ChatGPT utilizing demonstrations until the max length by a +13.98{\%} margin. Our code, the CoT Collection data, and model checkpoints are publicly available.

The CoT Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuning

emnlp-main.782

2305.14045

https://huggingface.co/papers/2305.14045

Poster

https://aclanthology.org/2023.emnlp-main.783.bib

https://aclanthology.org/2023.emnlp-main.783/

@inproceedings{choudhury-etal-2023-explaining, title = "Explaining Interactions Between Text Spans", author = "Ray Choudhury, Sagnik and Atanasova, Pepa and Augenstein, Isabelle", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.783", doi = "10.18653/v1/2023.emnlp-main.783", pages = "12709--12730", abstract = "Reasoning over spans of tokens from different parts of the input is essential for natural language understanding (NLU) tasks such as fact-checking (FC), machine reading comprehension (MRC) or natural language inference (NLI). However, existing highlight-based explanations primarily focus on identifying individual important features or interactions only between adjacent tokens or tuples of tokens. Most notably, there is a lack of annotations capturing the human decision-making process with respect to the necessary interactions for informed decision-making in such tasks. To bridge this gap, we introduce SpanEx, a multi-annotator dataset of human span interaction explanations for two NLU tasks: NLI and FC. We then investigate the decision-making processes of multiple fine-tuned large language models in terms of the employed connections between spans in separate parts of the input and compare them to the human reasoning processes. Finally, we present a novel community detection based unsupervised method to extract such interaction explanations. We make the code and the dataset available on [Github](https://github.com/copenlu/spanex). The dataset is also available on [Huggingface datasets](https://huggingface.co/datasets/copenlu/spanex).", }

Reasoning over spans of tokens from different parts of the input is essential for natural language understanding (NLU) tasks such as fact-checking (FC), machine reading comprehension (MRC) or natural language inference (NLI). However, existing highlight-based explanations primarily focus on identifying individual important features or interactions only between adjacent tokens or tuples of tokens. Most notably, there is a lack of annotations capturing the human decision-making process with respect to the necessary interactions for informed decision-making in such tasks. To bridge this gap, we introduce SpanEx, a multi-annotator dataset of human span interaction explanations for two NLU tasks: NLI and FC. We then investigate the decision-making processes of multiple fine-tuned large language models in terms of the employed connections between spans in separate parts of the input and compare them to the human reasoning processes. Finally, we present a novel community detection based unsupervised method to extract such interaction explanations. We make the code and the dataset available on [Github](https://github.com/copenlu/spanex). The dataset is also available on [Huggingface datasets](https://huggingface.co/datasets/copenlu/spanex).

Explaining Interactions Between Text Spans

emnlp-main.783

2310.13506

Oral

https://aclanthology.org/2023.emnlp-main.784.bib

https://aclanthology.org/2023.emnlp-main.784/

@inproceedings{qian-etal-2023-predictive, title = "Predictive Chemistry Augmented with Text Retrieval", author = "Qian, Yujie and Li, Zhening and Tu, Zhengkai and Coley, Connor and Barzilay, Regina", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.784", doi = "10.18653/v1/2023.emnlp-main.784", pages = "12731--12745", abstract = "This paper focuses on using natural language descriptions to enhance predictive models in the chemistry field. Conventionally, chemoinformatics models are trained with extensive structured data manually extracted from the literature. In this paper, we introduce TextReact, a novel method that directly augments predictive chemistry with texts retrieved from the literature. TextReact retrieves text descriptions relevant for a given chemical reaction, and then aligns them with the molecular representation of the reaction. This alignment is enhanced via an auxiliary masked LM objective incorporated in the predictor training. We empirically validate the framework on two chemistry tasks: reaction condition recommendation and one-step retrosynthesis. By leveraging text retrieval, TextReact significantly outperforms state-of-the-art chemoinformatics models trained solely on molecular data.", }

This paper focuses on using natural language descriptions to enhance predictive models in the chemistry field. Conventionally, chemoinformatics models are trained with extensive structured data manually extracted from the literature. In this paper, we introduce TextReact, a novel method that directly augments predictive chemistry with texts retrieved from the literature. TextReact retrieves text descriptions relevant for a given chemical reaction, and then aligns them with the molecular representation of the reaction. This alignment is enhanced via an auxiliary masked LM objective incorporated in the predictor training. We empirically validate the framework on two chemistry tasks: reaction condition recommendation and one-step retrosynthesis. By leveraging text retrieval, TextReact significantly outperforms state-of-the-art chemoinformatics models trained solely on molecular data.

Predictive Chemistry Augmented with Text Retrieval

emnlp-main.784

2312.04881

Oral

https://aclanthology.org/2023.emnlp-main.785.bib

https://aclanthology.org/2023.emnlp-main.785/

@inproceedings{qorib-ng-2023-system, title = "System Combination via Quality Estimation for Grammatical Error Correction", author = "Qorib, Muhammad Reza and Ng, Hwee Tou", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.785", doi = "10.18653/v1/2023.emnlp-main.785", pages = "12746--12759", abstract = "Quality estimation models have been developed to assess the corrections made by grammatical error correction (GEC) models when the reference or gold-standard corrections are not available. An ideal quality estimator can be utilized to combine the outputs of multiple GEC systems by choosing the best subset of edits from the union of all edits proposed by the GEC base systems. However, we found that existing GEC quality estimation models are not good enough in differentiating good corrections from bad ones, resulting in a low F0.5 score when used for system combination. In this paper, we propose GRECO, a new state-of-the-art quality estimation model that gives a better estimate of the quality of a corrected sentence, as indicated by having a higher correlation to the F0.5 score of a corrected sentence. It results in a combined GEC system with a higher F0.5 score. We also propose three methods for utilizing GEC quality estimation models for system combination with varying generality: model-agnostic, model-agnostic with voting bias, and model-dependent method. The combined GEC system outperforms the state of the art on the CoNLL-2014 test set and the BEA-2019 test set, achieving the highest F0.5 scores published to date.", }

Quality estimation models have been developed to assess the corrections made by grammatical error correction (GEC) models when the reference or gold-standard corrections are not available. An ideal quality estimator can be utilized to combine the outputs of multiple GEC systems by choosing the best subset of edits from the union of all edits proposed by the GEC base systems. However, we found that existing GEC quality estimation models are not good enough in differentiating good corrections from bad ones, resulting in a low F0.5 score when used for system combination. In this paper, we propose GRECO, a new state-of-the-art quality estimation model that gives a better estimate of the quality of a corrected sentence, as indicated by having a higher correlation to the F0.5 score of a corrected sentence. It results in a combined GEC system with a higher F0.5 score. We also propose three methods for utilizing GEC quality estimation models for system combination with varying generality: model-agnostic, model-agnostic with voting bias, and model-dependent method. The combined GEC system outperforms the state of the art on the CoNLL-2014 test set and the BEA-2019 test set, achieving the highest F0.5 scores published to date.

System Combination via Quality Estimation for Grammatical Error Correction

emnlp-main.785

2310.14947

https://huggingface.co/papers/2310.14947

Poster

https://aclanthology.org/2023.emnlp-main.786.bib

https://aclanthology.org/2023.emnlp-main.786/

@inproceedings{li-etal-2023-rethinking-negative, title = "Rethinking Negative Pairs in Code Search", author = "Li, Haochen and Zhou, Xin and Luu, Anh and Miao, Chunyan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.786", doi = "10.18653/v1/2023.emnlp-main.786", pages = "12760--12774", abstract = "Recently, contrastive learning has become a key component in fine-tuning code search models for software development efficiency and effectiveness. It pulls together positive code snippets while pushing negative samples away given search queries. Among contrastive learning, InfoNCE is the most widely used loss function due to its better performance. However, the following problems in negative samples of InfoNCE may deteriorate its representation learning: 1) The existence of false negative samples in large code corpora due to duplications. 2). The failure to explicitly differentiate between the potential relevance of negative samples. As an example, a bubble sorting algorithm example is less {``}negative{''} than a file saving function for the quick sorting algorithm query. In this paper, we tackle the above problems by proposing a simple yet effective Soft-InfoNCE loss that inserts weight terms into InfoNCE. In our proposed loss function, we apply three methods to estimate the weights of negative pairs and show that the vanilla InfoNCE loss is a special case of Soft-InfoNCE. Theoretically, we analyze the effects of Soft-InfoNCE on controlling the distribution of learnt code representations and on deducing a more precise mutual information estimation. We furthermore discuss the superiority of proposed loss functions with other design alternatives. Extensive experiments demonstrate the effectiveness of Soft-InfoNCE and weights estimation methods under state-of-the-art code search models on a large-scale public dataset consisting of six programming languages.", }

Recently, contrastive learning has become a key component in fine-tuning code search models for software development efficiency and effectiveness. It pulls together positive code snippets while pushing negative samples away given search queries. Among contrastive learning, InfoNCE is the most widely used loss function due to its better performance. However, the following problems in negative samples of InfoNCE may deteriorate its representation learning: 1) The existence of false negative samples in large code corpora due to duplications. 2). The failure to explicitly differentiate between the potential relevance of negative samples. As an example, a bubble sorting algorithm example is less {``}negative{''} than a file saving function for the quick sorting algorithm query. In this paper, we tackle the above problems by proposing a simple yet effective Soft-InfoNCE loss that inserts weight terms into InfoNCE. In our proposed loss function, we apply three methods to estimate the weights of negative pairs and show that the vanilla InfoNCE loss is a special case of Soft-InfoNCE. Theoretically, we analyze the effects of Soft-InfoNCE on controlling the distribution of learnt code representations and on deducing a more precise mutual information estimation. We furthermore discuss the superiority of proposed loss functions with other design alternatives. Extensive experiments demonstrate the effectiveness of Soft-InfoNCE and weights estimation methods under state-of-the-art code search models on a large-scale public dataset consisting of six programming languages.

Rethinking Negative Pairs in Code Search

emnlp-main.786

2310.08069

Poster

https://aclanthology.org/2023.emnlp-main.787.bib

https://aclanthology.org/2023.emnlp-main.787/

@inproceedings{zhu-etal-2023-question, title = "Question Answering as Programming for Solving Time-Sensitive Questions", author = "Zhu, Xinyu and Yang, Cheng and Chen, Bei and Li, Siheng and Lou, Jian-Guang and Yang, Yujiu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.787", doi = "10.18653/v1/2023.emnlp-main.787", pages = "12775--12790", abstract = "Question answering plays a pivotal role in human daily life because it involves our acquisition of knowledge about the world. However, due to the dynamic and ever-changing nature of real-world facts, the answer can be completely different when the time constraint in the question changes. Recently, Large Language Models (LLMs) have shown remarkable intelligence in question answering, while our experiments reveal that the aforementioned problems still pose a significant challenge to existing LLMs. This can be attributed to the LLMs{'} inability to perform rigorous reasoning based on surface-level text semantics. To overcome this limitation, rather than requiring LLMs to directly answer the question, we propose a novel approach where we reframe the $\textbf{Q}$uestion $\textbf{A}$nswering task $\textbf{a}$s $\textbf{P}$rogramming ($\textbf{QAaP}$). Concretely, by leveraging modern LLMs{'} superior capability in understanding both natural language and programming language, we endeavor to harness LLMs to represent diversely expressed text as well-structured code and select the best matching answer from multiple candidates through programming. We evaluate our QAaP framework on several time-sensitive question answering datasets and achieve decent improvement, up to 14.5{\%} over strong baselines.", }

Question answering plays a pivotal role in human daily life because it involves our acquisition of knowledge about the world. However, due to the dynamic and ever-changing nature of real-world facts, the answer can be completely different when the time constraint in the question changes. Recently, Large Language Models (LLMs) have shown remarkable intelligence in question answering, while our experiments reveal that the aforementioned problems still pose a significant challenge to existing LLMs. This can be attributed to the LLMs{'} inability to perform rigorous reasoning based on surface-level text semantics. To overcome this limitation, rather than requiring LLMs to directly answer the question, we propose a novel approach where we reframe the $\textbf{Q}$uestion $\textbf{A}$nswering task $\textbf{a}$s $\textbf{P}$rogramming ($\textbf{QAaP}$). Concretely, by leveraging modern LLMs{'} superior capability in understanding both natural language and programming language, we endeavor to harness LLMs to represent diversely expressed text as well-structured code and select the best matching answer from multiple candidates through programming. We evaluate our QAaP framework on several time-sensitive question answering datasets and achieve decent improvement, up to 14.5{\%} over strong baselines.

Question Answering as Programming for Solving Time-Sensitive Questions

emnlp-main.787

2305.14221

https://huggingface.co/papers/2305.14221

Poster

https://aclanthology.org/2023.emnlp-main.788.bib

https://aclanthology.org/2023.emnlp-main.788/

@inproceedings{du-etal-2023-joint, title = "Joint Geometrical and Statistical Domain Adaptation for Cross-domain Code Vulnerability Detection", author = "Du, Qianjin and Zhou, Shiji and Kuang, Xiaohui and Zhao, Gang and Zhai, Jidong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.788", doi = "10.18653/v1/2023.emnlp-main.788", pages = "12791--12800", abstract = "In code vulnerability detection tasks, a detector trained on a label-rich source domain fails to provide accurate prediction on new or unseen target domains due to the lack of labeled training data on target domains. Previous studies mainly utilize domain adaptation to perform cross-domain vulnerability detection. But they ignore the negative effect of private semantic characteristics of the target domain for domain alignment, which easily causes the problem of negative transfer. In addition, these methods forcibly reduce the distribution discrepancy between domains and do not take into account the interference of irrelevant target instances for distributional domain alignment, which leads to the problem of excessive alignment. To address the above issues, we propose a novel cross-domain code vulnerability detection framework named MNCRI. Specifically, we introduce mutual nearest neighbor contrastive learning to align the source domain and target domain geometrically, which could align the common semantic characteristics of two domains and separate out the private semantic characteristics of each domain. Furthermore, we introduce an instance re-weighting scheme to alleviate the problem of excessive alignment. This scheme dynamically assign different weights to instances, reducing the contribution of irrelevant instances so as to achieve better domain alignment. Finally, extensive experiments demonstrate that MNCRI significantly outperforms state-of-the-art cross-domain code vulnerability detection methods by a large margin.", }

In code vulnerability detection tasks, a detector trained on a label-rich source domain fails to provide accurate prediction on new or unseen target domains due to the lack of labeled training data on target domains. Previous studies mainly utilize domain adaptation to perform cross-domain vulnerability detection. But they ignore the negative effect of private semantic characteristics of the target domain for domain alignment, which easily causes the problem of negative transfer. In addition, these methods forcibly reduce the distribution discrepancy between domains and do not take into account the interference of irrelevant target instances for distributional domain alignment, which leads to the problem of excessive alignment. To address the above issues, we propose a novel cross-domain code vulnerability detection framework named MNCRI. Specifically, we introduce mutual nearest neighbor contrastive learning to align the source domain and target domain geometrically, which could align the common semantic characteristics of two domains and separate out the private semantic characteristics of each domain. Furthermore, we introduce an instance re-weighting scheme to alleviate the problem of excessive alignment. This scheme dynamically assign different weights to instances, reducing the contribution of irrelevant instances so as to achieve better domain alignment. Finally, extensive experiments demonstrate that MNCRI significantly outperforms state-of-the-art cross-domain code vulnerability detection methods by a large margin.

Joint Geometrical and Statistical Domain Adaptation for Cross-domain Code Vulnerability Detection

emnlp-main.788

Poster

https://aclanthology.org/2023.emnlp-main.789.bib

https://aclanthology.org/2023.emnlp-main.789/

@inproceedings{chen-etal-2023-revisiting-sparse, title = "Revisiting Sparse Retrieval for Few-shot Entity Linking", author = "Chen, Yulin and Xu, Zhenran and Hu, Baotian and Zhang, Min", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.789", doi = "10.18653/v1/2023.emnlp-main.789", pages = "12801--12806", abstract = "Entity linking aims to link ambiguous mentions to their corresponding entities in a knowledge base. One of the key challenges comes from insufficient labeled data for specific domains. Although dense retrievers have achieved excellent performance on several benchmarks, their performance decreases significantly when only a limited amount of in-domain labeled data is available. In such few-shot setting, we revisit the sparse retrieval method, and propose an ELECTRA-based keyword extractor to denoise the mention context and construct a better query expression. For training the extractor, we propose a distant supervision method to automatically generate training data based on overlapping tokens between mention contexts and entity descriptions. Experimental results on the ZESHEL dataset demonstrate that the proposed method outperforms state-of-the-art models by a significant margin across all test domains, showing the effectiveness of keyword-enhanced sparse retrieval.", }

Entity linking aims to link ambiguous mentions to their corresponding entities in a knowledge base. One of the key challenges comes from insufficient labeled data for specific domains. Although dense retrievers have achieved excellent performance on several benchmarks, their performance decreases significantly when only a limited amount of in-domain labeled data is available. In such few-shot setting, we revisit the sparse retrieval method, and propose an ELECTRA-based keyword extractor to denoise the mention context and construct a better query expression. For training the extractor, we propose a distant supervision method to automatically generate training data based on overlapping tokens between mention contexts and entity descriptions. Experimental results on the ZESHEL dataset demonstrate that the proposed method outperforms state-of-the-art models by a significant margin across all test domains, showing the effectiveness of keyword-enhanced sparse retrieval.

Revisiting Sparse Retrieval for Few-shot Entity Linking

emnlp-main.789

2310.12444

Poster

https://aclanthology.org/2023.emnlp-main.790.bib

https://aclanthology.org/2023.emnlp-main.790/

@inproceedings{agrawal-carpuat-2023-controlling, title = "Controlling Pre-trained Language Models for Grade-Specific Text Simplification", author = "Agrawal, Sweta and Carpuat, Marine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.790", doi = "10.18653/v1/2023.emnlp-main.790", pages = "12807--12819", abstract = "Text simplification systems rewrite text to make it more readable while preserving its content. However, what makes a text easy to read depends on the intended readers. Recent work has shown that pre-trained language models can simplify text using a wealth of techniques to control output simplicity, ranging from specifying only the desired reading grade level, to directly specifying low-level edit operations. Yet it remains unclear how to set these control parameters in practice. Existing approaches set them at the corpus level, disregarding the complexity of individual inputs and considering only one level of output complexity. In this work, we conduct an empirical study to understand how different control mechanisms impact the adequacy and simplicity of text simplification systems. Based on these insights, we introduce a simple method that predicts the edit operations required for simplifying a text for a specific grade level on an instance-per-instance basis. This approach improves the quality of the simplified outputs over corpus-level search-based heuristics.", }

Text simplification systems rewrite text to make it more readable while preserving its content. However, what makes a text easy to read depends on the intended readers. Recent work has shown that pre-trained language models can simplify text using a wealth of techniques to control output simplicity, ranging from specifying only the desired reading grade level, to directly specifying low-level edit operations. Yet it remains unclear how to set these control parameters in practice. Existing approaches set them at the corpus level, disregarding the complexity of individual inputs and considering only one level of output complexity. In this work, we conduct an empirical study to understand how different control mechanisms impact the adequacy and simplicity of text simplification systems. Based on these insights, we introduce a simple method that predicts the edit operations required for simplifying a text for a specific grade level on an instance-per-instance basis. This approach improves the quality of the simplified outputs over corpus-level search-based heuristics.

Controlling Pre-trained Language Models for Grade-Specific Text Simplification

emnlp-main.790

2305.14993

Poster

https://aclanthology.org/2023.emnlp-main.791.bib

https://aclanthology.org/2023.emnlp-main.791/

@inproceedings{zhang-etal-2023-clevr, title = "{CLEVR}-Implicit: A Diagnostic Dataset for Implicit Reasoning in Referring Expression Comprehension", author = "Zhang, Jingwei and Wu, Xin and Cai, Yi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.791", doi = "10.18653/v1/2023.emnlp-main.791", pages = "12820--12830", abstract = "Recently, pre-trained vision-language (VL) models have achieved remarkable success in various cross-modal tasks, including referring expression comprehension (REC). These models are pre-trained on the large-scale image-text pairs to learn the alignment between words in textual descriptions and objects in the corresponding images and then fine-tuned on downstream tasks. However, the performance of VL models is hindered when dealing with implicit text, which describes objects through comparisons between two or more objects rather than explicitly mentioning them. This is because the models struggle to align the implicit text with the objects in the images. To address the challenge, we introduce CLEVR-Implicit, a dataset consisting of synthetic images and corresponding two types of implicit text for the REC task. Additionally, to enhance the performance of VL models on implicit text, we propose a method called Transforming Implicit text into Explicit text (TIE), which enables VL models to reason with the implicit text. TIE consists of two modules: (1) the prompt design module builds prompts for implicit text by adding masked tokens, and (2) the cloze procedure module fine-tunes the prompts by utilizing masked language modeling (MLM) to predict the explicit words with the implicit prompts. Experimental results on our dataset demonstrate a significant improvement of 37.94{\%} in the performance of VL models on implicit text after employing our TIE method.", }

Recently, pre-trained vision-language (VL) models have achieved remarkable success in various cross-modal tasks, including referring expression comprehension (REC). These models are pre-trained on the large-scale image-text pairs to learn the alignment between words in textual descriptions and objects in the corresponding images and then fine-tuned on downstream tasks. However, the performance of VL models is hindered when dealing with implicit text, which describes objects through comparisons between two or more objects rather than explicitly mentioning them. This is because the models struggle to align the implicit text with the objects in the images. To address the challenge, we introduce CLEVR-Implicit, a dataset consisting of synthetic images and corresponding two types of implicit text for the REC task. Additionally, to enhance the performance of VL models on implicit text, we propose a method called Transforming Implicit text into Explicit text (TIE), which enables VL models to reason with the implicit text. TIE consists of two modules: (1) the prompt design module builds prompts for implicit text by adding masked tokens, and (2) the cloze procedure module fine-tunes the prompts by utilizing masked language modeling (MLM) to predict the explicit words with the implicit prompts. Experimental results on our dataset demonstrate a significant improvement of 37.94{\%} in the performance of VL models on implicit text after employing our TIE method.

CLEVR-Implicit: A Diagnostic Dataset for Implicit Reasoning in Referring Expression Comprehension

emnlp-main.791

Poster

https://aclanthology.org/2023.emnlp-main.792.bib

https://aclanthology.org/2023.emnlp-main.792/

@inproceedings{burger-etal-2023-explanations, title = "{``}Are Your Explanations Reliable?{''} Investigating the Stability of {LIME} in Explaining Text Classifiers by Marrying {XAI} and Adversarial Attack", author = "Burger, Christopher and Chen, Lingwei and Le, Thai", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.792", doi = "10.18653/v1/2023.emnlp-main.792", pages = "12831--12844", abstract = "LIME has emerged as one of the most commonly referenced tools in explainable AI (XAI) frameworks that is integrated into critical machine learning applications (e.g., healthcare and finance). However, its stability remains little explored, especially in the context of text data, due to the unique text-space constraints. To address these challenges, in this paper, we first evaluate the inherent instability of LIME on text data to establish a baseline, and then propose a novel algorithm XAIFooler to perturb text inputs and manipulate explanations that casts investigation on the stability of LIME as a text perturbation optimization problem. XAIFooler conforms to the constraints to preserve text semantics and original prediction with small perturbations, and introduces Rank-biased Overlap (RBO) as a key part to guide the optimization of XAIFooler that satisfies all the requirements for explanation similarity measure. Extensive experiments on real-world text datasets demonstrate that XAIFooler significantly outperforms all baselines by large margins in its ability to manipulate LIME{'}s explanations with high semantic preservability.", }

LIME has emerged as one of the most commonly referenced tools in explainable AI (XAI) frameworks that is integrated into critical machine learning applications (e.g., healthcare and finance). However, its stability remains little explored, especially in the context of text data, due to the unique text-space constraints. To address these challenges, in this paper, we first evaluate the inherent instability of LIME on text data to establish a baseline, and then propose a novel algorithm XAIFooler to perturb text inputs and manipulate explanations that casts investigation on the stability of LIME as a text perturbation optimization problem. XAIFooler conforms to the constraints to preserve text semantics and original prediction with small perturbations, and introduces Rank-biased Overlap (RBO) as a key part to guide the optimization of XAIFooler that satisfies all the requirements for explanation similarity measure. Extensive experiments on real-world text datasets demonstrate that XAIFooler significantly outperforms all baselines by large margins in its ability to manipulate LIME{'}s explanations with high semantic preservability.

“Are Your Explanations Reliable?” Investigating the Stability of LIME in Explaining Text Classifiers by Marrying XAI and Adversarial Attack

emnlp-main.792

Poster

https://aclanthology.org/2023.emnlp-main.793.bib

https://aclanthology.org/2023.emnlp-main.793/

@inproceedings{almasian-etal-2023-cqe, title = "{CQE}: A Comprehensive Quantity Extractor", author = {Almasian, Satya and Kazakova, Vivian and G{\"o}ldner, Philipp and Gertz, Michael}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.793", doi = "10.18653/v1/2023.emnlp-main.793", pages = "12845--12859", abstract = "Quantities are essential in documents to describe factual information. They are ubiquitous in application domains such as finance, business, medicine, and science in general. Compared to other information extraction approaches, interestingly only a few works exist that describe methods for a proper extraction and representation of quantities in text. In this paper, we present such a comprehensive quantity extraction framework from text data. It efficiently detects combinations of values and units, the behavior of a quantity (e.g., rising or falling), and the concept a quantity is associated with. Our framework makes use of dependency parsing and a dictionary of units, and it provides for a proper normalization and standardization of detected quantities. Using a novel dataset for evaluation, we show that our open source framework outperforms other systems and {--} to the best of our knowledge {--} is the first to detect concepts associated with identified quantities. The code and data underlying our framework are available at https://github.com/vivkaz/CQE.", }

Quantities are essential in documents to describe factual information. They are ubiquitous in application domains such as finance, business, medicine, and science in general. Compared to other information extraction approaches, interestingly only a few works exist that describe methods for a proper extraction and representation of quantities in text. In this paper, we present such a comprehensive quantity extraction framework from text data. It efficiently detects combinations of values and units, the behavior of a quantity (e.g., rising or falling), and the concept a quantity is associated with. Our framework makes use of dependency parsing and a dictionary of units, and it provides for a proper normalization and standardization of detected quantities. Using a novel dataset for evaluation, we show that our open source framework outperforms other systems and {--} to the best of our knowledge {--} is the first to detect concepts associated with identified quantities. The code and data underlying our framework are available at https://github.com/vivkaz/CQE.

CQE: A Comprehensive Quantity Extractor

emnlp-main.793

2305.08853

Poster

https://aclanthology.org/2023.emnlp-main.794.bib

https://aclanthology.org/2023.emnlp-main.794/

@inproceedings{ren-etal-2023-context, title = "Context Compression for Auto-regressive Transformers with Sentinel Tokens", author = "Ren, Siyu and Jia, Qi and Zhu, Kenny", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.794", doi = "10.18653/v1/2023.emnlp-main.794", pages = "12860--12867", abstract = "The quadratic complexity of the attention module makes it gradually become the bulk of compute in Transformer-based LLMs during generation. Moreover, the excessive key-value cache that arises when dealing with long inputs also brings severe issues on memory footprint and inference latency. In this work, we propose a plug-and-play approach that is able to incrementally compress the intermediate activation of a specified span of tokens into compact ones, thereby reducing both memory and computational cost when processing subsequent context. Experiments on both in-domain language modeling and zero-shot open-ended document generation demonstrate the advantage of our approach over sparse attention baselines in terms of fluency, n-gram matching, and semantic similarity. At last, we comprehensively profile the benefit of context compression on improving the system throughout. Code is available at \url{https://github.com/DRSY/KV_Compression}.", }

The quadratic complexity of the attention module makes it gradually become the bulk of compute in Transformer-based LLMs during generation. Moreover, the excessive key-value cache that arises when dealing with long inputs also brings severe issues on memory footprint and inference latency. In this work, we propose a plug-and-play approach that is able to incrementally compress the intermediate activation of a specified span of tokens into compact ones, thereby reducing both memory and computational cost when processing subsequent context. Experiments on both in-domain language modeling and zero-shot open-ended document generation demonstrate the advantage of our approach over sparse attention baselines in terms of fluency, n-gram matching, and semantic similarity. At last, we comprehensively profile the benefit of context compression on improving the system throughout. Code is available at \url{https://github.com/DRSY/KV_Compression}.

Context Compression for Auto-regressive Transformers with Sentinel Tokens

emnlp-main.794

2310.08152

https://huggingface.co/papers/2310.08152

Poster

https://aclanthology.org/2023.emnlp-main.795.bib

https://aclanthology.org/2023.emnlp-main.795/

@inproceedings{chen-etal-2023-unified, title = "A Unified View of Evaluation Metrics for Structured Prediction", author = "Chen, Yunmo and Gantt, William and Chen, Tongfei and White, Aaron and Van Durme, Benjamin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.795", doi = "10.18653/v1/2023.emnlp-main.795", pages = "12868--12882", abstract = "We present a conceptual framework that unifies a variety of evaluation metrics for different structured prediction tasks (e.g. event and relation extraction, syntactic and semantic parsing). Our framework requires representing the outputs of these tasks as objects of certain data types, and derives metrics through matching of common substructures, possibly followed by normalization. We demonstrate how commonly used metrics for a number of tasks can be succinctly expressed by this framework, and show that new metrics can be naturally derived in a bottom-up way based on an output structure. We release a library that enables this derivation to create new metrics. Finally, we consider how specific characteristics of tasks motivate metric design decisions, and suggest possible modifications to existing metrics in line with those motivations.", }

We present a conceptual framework that unifies a variety of evaluation metrics for different structured prediction tasks (e.g. event and relation extraction, syntactic and semantic parsing). Our framework requires representing the outputs of these tasks as objects of certain data types, and derives metrics through matching of common substructures, possibly followed by normalization. We demonstrate how commonly used metrics for a number of tasks can be succinctly expressed by this framework, and show that new metrics can be naturally derived in a bottom-up way based on an output structure. We release a library that enables this derivation to create new metrics. Finally, we consider how specific characteristics of tasks motivate metric design decisions, and suggest possible modifications to existing metrics in line with those motivations.

A Unified View of Evaluation Metrics for Structured Prediction

emnlp-main.795

2310.13793

Poster

https://aclanthology.org/2023.emnlp-main.796.bib

https://aclanthology.org/2023.emnlp-main.796/

@inproceedings{tsur-tulpan-2023-deeper, title = "A Deeper (Autoregressive) Approach to Non-Convergent Discourse Parsing", author = "Tsur, Oren and Tulpan, Yoav", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.796", doi = "10.18653/v1/2023.emnlp-main.796", pages = "12883--12895", abstract = "Online social platforms provide a bustling arena for information-sharing and for multi-party discussions. Various frameworks for dialogic discourse parsing were developed and used for the processing of discussions and for predicting the productivity of a dialogue. However, most of these frameworks are not suitable for the analysis of contentious discussions that are commonplace in many online platforms. A novel multi-label scheme for contentious dialog parsing was recently introduced by Zakharov et al. (2021). While the schema is well developed, the computational approach they provide is both naive and inefficient, as a different model (architecture) using a different representation of the input, is trained for each of the 31 tags in the annotation scheme. Moreover, all their models assume full knowledge of label collocations and context, which is unlikely in any realistic setting. In this work, we present a unified model for Non-Convergent Discourse Parsing that does not require any additional input other than the previous dialog utterances. We fine-tuned a RoBERTa backbone, combining embeddings of the utterance, the context and the labels through GRN layers and an asymmetric loss function. Overall, our model achieves results comparable with SOTA, without using label collocation and without training a unique architecture/model for each label. Our proposed architecture makes the labeling feasible at large scale, promoting the development of tools that deepen our understanding of discourse dynamics.", }

Online social platforms provide a bustling arena for information-sharing and for multi-party discussions. Various frameworks for dialogic discourse parsing were developed and used for the processing of discussions and for predicting the productivity of a dialogue. However, most of these frameworks are not suitable for the analysis of contentious discussions that are commonplace in many online platforms. A novel multi-label scheme for contentious dialog parsing was recently introduced by Zakharov et al. (2021). While the schema is well developed, the computational approach they provide is both naive and inefficient, as a different model (architecture) using a different representation of the input, is trained for each of the 31 tags in the annotation scheme. Moreover, all their models assume full knowledge of label collocations and context, which is unlikely in any realistic setting. In this work, we present a unified model for Non-Convergent Discourse Parsing that does not require any additional input other than the previous dialog utterances. We fine-tuned a RoBERTa backbone, combining embeddings of the utterance, the context and the labels through GRN layers and an asymmetric loss function. Overall, our model achieves results comparable with SOTA, without using label collocation and without training a unique architecture/model for each label. Our proposed architecture makes the labeling feasible at large scale, promoting the development of tools that deepen our understanding of discourse dynamics.

A Deeper (Autoregressive) Approach to Non-Convergent Discourse Parsing

emnlp-main.796

2305.12510

Poster

https://aclanthology.org/2023.emnlp-main.797.bib

https://aclanthology.org/2023.emnlp-main.797/

@inproceedings{wahle-etal-2023-cite, title = "We are Who We Cite: Bridges of Influence Between Natural Language Processing and Other Academic Fields", author = "Wahle, Jan Philip and Ruas, Terry and Abdalla, Mohamed and Gipp, Bela and Mohammad, Saif", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.797", doi = "10.18653/v1/2023.emnlp-main.797", pages = "12896--12913", abstract = "Natural Language Processing (NLP) is poised to substantially influence the world. However, significant progress comes hand-in-hand with substantial risks. Addressing them requires broad engagement with various fields of study. Yet, little empirical work examines the state of such engagement (past or current). In this paper, we quantify the degree of influence between 23 fields of study and NLP (on each other). We analyzed {\textasciitilde}77k NLP papers, {\textasciitilde}3.1m citations from NLP papers to other papers, and {\textasciitilde}1.8m citations from other papers to NLP papers. We show that, unlike most fields, the cross-field engagement of NLP, measured by our proposed Citation Field Diversity Index (CFDI), has declined from 0.58 in 1980 to 0.31 in 2022 (an all-time low). In addition, we find that NLP has grown more insular{---}citing increasingly more NLP papers and having fewer papers that act as bridges between fields. NLP citations are dominated by computer science; Less than 8{\%} of NLP citations are to linguistics, and less than 3{\%} are to math and psychology. These findings underscore NLP{'}s urgent need to reflect on its engagement with various fields.", }

Natural Language Processing (NLP) is poised to substantially influence the world. However, significant progress comes hand-in-hand with substantial risks. Addressing them requires broad engagement with various fields of study. Yet, little empirical work examines the state of such engagement (past or current). In this paper, we quantify the degree of influence between 23 fields of study and NLP (on each other). We analyzed {\textasciitilde}77k NLP papers, {\textasciitilde}3.1m citations from NLP papers to other papers, and {\textasciitilde}1.8m citations from other papers to NLP papers. We show that, unlike most fields, the cross-field engagement of NLP, measured by our proposed Citation Field Diversity Index (CFDI), has declined from 0.58 in 1980 to 0.31 in 2022 (an all-time low). In addition, we find that NLP has grown more insular{---}citing increasingly more NLP papers and having fewer papers that act as bridges between fields. NLP citations are dominated by computer science; Less than 8{\%} of NLP citations are to linguistics, and less than 3{\%} are to math and psychology. These findings underscore NLP{'}s urgent need to reflect on its engagement with various fields.

We are Who We Cite: Bridges of Influence Between Natural Language Processing and Other Academic Fields

emnlp-main.797

2310.14870

Oral

https://aclanthology.org/2023.emnlp-main.798.bib

https://aclanthology.org/2023.emnlp-main.798/

@inproceedings{deutsch-etal-2023-ties, title = "Ties Matter: Meta-Evaluating Modern Metrics with Pairwise Accuracy and Tie Calibration", author = "Deutsch, Daniel and Foster, George and Freitag, Markus", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.798", doi = "10.18653/v1/2023.emnlp-main.798", pages = "12914--12929", abstract = "Kendall{'}s tau is frequently used to meta-evaluate how well machine translation (MT) evaluation metrics score individual translations. Its focus on pairwise score comparisons is intuitive but raises the question of how ties should be handled, a gray area that has motivated different variants in the literature. We demonstrate that, in settings like modern MT meta-evaluation, existing variants have weaknesses arising from their handling of ties, and in some situations can even be gamed. We propose instead to meta-evaluate metrics with a version of pairwise accuracy that gives metrics credit for correctly predicting ties, in combination with a tie calibration procedure that automatically introduces ties into metric scores, enabling fair comparison between metrics that do and do not predict ties. We argue and provide experimental evidence that these modifications lead to fairer ranking-based assessments of metric performance.", }

Kendall{'}s tau is frequently used to meta-evaluate how well machine translation (MT) evaluation metrics score individual translations. Its focus on pairwise score comparisons is intuitive but raises the question of how ties should be handled, a gray area that has motivated different variants in the literature. We demonstrate that, in settings like modern MT meta-evaluation, existing variants have weaknesses arising from their handling of ties, and in some situations can even be gamed. We propose instead to meta-evaluate metrics with a version of pairwise accuracy that gives metrics credit for correctly predicting ties, in combination with a tie calibration procedure that automatically introduces ties into metric scores, enabling fair comparison between metrics that do and do not predict ties. We argue and provide experimental evidence that these modifications lead to fairer ranking-based assessments of metric performance.

Ties Matter: Meta-Evaluating Modern Metrics with Pairwise Accuracy and Tie Calibration

emnlp-main.798

2305.14324

Poster

https://aclanthology.org/2023.emnlp-main.799.bib

https://aclanthology.org/2023.emnlp-main.799/

@inproceedings{kim-etal-2023-soda, title = "{SODA}: Million-scale Dialogue Distillation with Social Commonsense Contextualization", author = "Kim, Hyunwoo and Hessel, Jack and Jiang, Liwei and West, Peter and Lu, Ximing and Yu, Youngjae and Zhou, Pei and Bras, Ronan and Alikhani, Malihe and Kim, Gunhee and Sap, Maarten and Choi, Yejin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.799", doi = "10.18653/v1/2023.emnlp-main.799", pages = "12930--12949", abstract = "Data scarcity has been a long standing issue in the field of open-domain social dialogue. To quench this thirst, we present SODA: the first publicly available, million-scale high-quality social dialogue dataset. By contextualizing social commonsense knowledge from a knowledge graph, we are able to distill an exceptionally broad spectrum of social interactions from a large language model. Human evaluation shows that conversations in SODA are more consistent, specific, and (surprisingly) natural than those in prior human-authored datasets. Using SODA, we train COSMO: a generalizable conversation model that is significantly more natural and consistent on unseen datasets than best-performing conversation models (e.g., GODEL, BlenderBot-1, Koala, Vicuna). Experiments reveal COSMO is sometimes even preferred to the original human-written gold responses. Additionally, our results shed light on the distinction between knowledge-enriched conversations and natural social chitchats. We plan to make our data, model, and code public.", }

Data scarcity has been a long standing issue in the field of open-domain social dialogue. To quench this thirst, we present SODA: the first publicly available, million-scale high-quality social dialogue dataset. By contextualizing social commonsense knowledge from a knowledge graph, we are able to distill an exceptionally broad spectrum of social interactions from a large language model. Human evaluation shows that conversations in SODA are more consistent, specific, and (surprisingly) natural than those in prior human-authored datasets. Using SODA, we train COSMO: a generalizable conversation model that is significantly more natural and consistent on unseen datasets than best-performing conversation models (e.g., GODEL, BlenderBot-1, Koala, Vicuna). Experiments reveal COSMO is sometimes even preferred to the original human-written gold responses. Additionally, our results shed light on the distinction between knowledge-enriched conversations and natural social chitchats. We plan to make our data, model, and code public.

SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization

emnlp-main.799

2212.10465

https://huggingface.co/papers/2212.10465

Oral

https://aclanthology.org/2023.emnlp-main.800.bib

https://aclanthology.org/2023.emnlp-main.800/

@inproceedings{hu-etal-2023-multi, title = "Multi-view Contrastive Learning for Entity Typing over Knowledge Graphs", author = "Hu, Zhiwei and Basulto, Victor and Xiang, Zhiliang and Li, Ru and Pan, Jeff", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.800", doi = "10.18653/v1/2023.emnlp-main.800", pages = "12950--12963", abstract = "Knowledge graph entity typing (KGET) aims at inferring plausible types of entities in knowledge graphs. Existing approaches to KGET focus on how to better encode the knowledge provided by the neighbors and types of an entity into its representation. However, they ignore the semantic knowledge provided by the way in which types can be clustered together. In this paper, we propose a novel method called Multi-view Contrastive Learning for knowledge graph Entity Typing MCLET, which effectively encodes the coarse-grained knowledge provided by clusters into entity and type embeddings. MCLET is composed of three modules: i) Multi-view Generation and Encoder module, which encodes structured information from entity-type, entity-cluster and cluster-type views; ii) Cross-view Contrastive Learning module, which encourages different views to collaboratively improve view-specific representations of entities and types; iii) Entity Typing Prediction module, which integrates multi-head attention and a Mixture-of-Experts strategy to infer missing entity types. Extensive experiments show the strong performance of MCLET compared to the state-of-the-art", }

Knowledge graph entity typing (KGET) aims at inferring plausible types of entities in knowledge graphs. Existing approaches to KGET focus on how to better encode the knowledge provided by the neighbors and types of an entity into its representation. However, they ignore the semantic knowledge provided by the way in which types can be clustered together. In this paper, we propose a novel method called Multi-view Contrastive Learning for knowledge graph Entity Typing MCLET, which effectively encodes the coarse-grained knowledge provided by clusters into entity and type embeddings. MCLET is composed of three modules: i) Multi-view Generation and Encoder module, which encodes structured information from entity-type, entity-cluster and cluster-type views; ii) Cross-view Contrastive Learning module, which encourages different views to collaboratively improve view-specific representations of entities and types; iii) Entity Typing Prediction module, which integrates multi-head attention and a Mixture-of-Experts strategy to infer missing entity types. Extensive experiments show the strong performance of MCLET compared to the state-of-the-art

Multi-view Contrastive Learning for Entity Typing over Knowledge Graphs

emnlp-main.800

2310.12008

https://huggingface.co/papers/2310.12008

Poster