Kundyzka/kaz-roberta-conversational-informatics

Description

This model was developed by Kundyz Maksutova, PhD Candidate, as part of research on improving question-answering systems in the Kazakh language. It is a fine-tuned version of kz-transformers/kaz-roberta-conversational using the Kundyzka/informatics_kaz dataset. The model is optimized for answering questions in Kazakh, with a primary focus on computer science and related fields.

Key Features:

• Developer: Kundyz Maksutova, PhD Candidate
• Base Model: kz-transformers/kaz-roberta-conversational
• Dataset: Kundyzka/informatics_kaz
• Language: Kazakh (kk)
• Task: Question Answering (pipeline_tag: question-answering)
• Library: adapter-transformers

Performance:

The model achieves the following performance metrics, highlighting its improvement after fine-tuning:

• Before Training:
  • F1 Score: 17.797
  • Exact Match (EM): 7.662
• After Training:
  • F1 Score: 67.788
  • Exact Match (EM): 51.428

These metrics were evaluated on the Kundyzka/informatics_kaz dataset, demonstrating a significant improvement in performance and reliability for domain-specific questions.

Intended Use:

This model is designed to handle natural language questions in the Kazakh language. It is particularly well-suited for:

• Educational Platforms: Assisting students with questions in computer science.
• Research Projects: Facilitating studies and experiments in Kazakh natural language processing.
• AI Applications: Powering chatbots and intelligent systems requiring accurate and domain-specific answers.

Limitations:

• Domain Dependency: The model is fine-tuned for computer science topics, and performance may degrade on unrelated queries.
• Bias: The training dataset may introduce biases that could affect the model’s responses.
• Language: The model supports only the Kazakh language and is not designed for multilingual use.

Tags:

• computerscience
• question-answering
• Kazakh
• adapter-transformers

This model contributes to advancing natural language processing for low-resource languages like Kazakh, with a focus on computer science applications. For further details, fine-tuning guidelines, or customization, refer to the model repository.