macbert4csc_v2
概述(macbert4csc_v2)
- macro-correct, 中文拼写纠错CSC测评(文本纠错), 权重使用
- 项目地址在https://github.com/yongzhuo/macro-correct
- 本模型权重为macbert4csc_v2, 使用macbert4csc架构(pycorrector版本), 其特点是在BertForMaskedLM后新加一个分支用于错误检测任务(分类任务, 不交互);
- 训练时使用了MFT(动态mask 0.2的非错误tokens), 同时det_loss的权重为0.3;
- 推理时舍弃了macbert后面的部分(det-layer);
- 如何使用: 1.使用transformers调用; 2.使用macro-correct项目调用; 详情见三、调用(Usage);
目录
一、测评(Test)
1.1 测评数据来源
地址为Macropodus/csc_eval_public, 所有训练数据均来自公网或开源数据, 训练数据为1千万左右, 混淆词典较大;
1.gen_de3.json(5545): '的地得'纠错, 由人民日报/学习强国/chinese-poetry等高质量数据人工生成;
2.lemon_v2.tet.json(1053): relm论文提出的数据, 多领域拼写纠错数据集(7个领域), ; 包括game(GAM), encyclopedia (ENC), contract (COT), medical care(MEC), car (CAR), novel (NOV), and news (NEW)等领域;
3.acc_rmrb.tet.json(4636): 来自NER-199801(人民日报高质量语料);
4.acc_xxqg.tet.json(5000): 来自学习强国网站的高质量语料;
5.gen_passage.tet.json(10000): 源数据为qwen生成的好词好句, 由几乎所有的开源数据汇总的混淆词典生成;
6.textproof.tet.json(1447): NLP竞赛数据, TextProofreadingCompetition;
7.gen_xxqg.tet.json(5000): 源数据为学习强国网站的高质量语料, 由几乎所有的开源数据汇总的混淆词典生成;
8.faspell.dev.json(1000): 视频字幕通过OCR后获取的数据集; 来自爱奇艺的论文faspell;
9.lomo_tet.json(5000): 主要为音似中文拼写纠错数据集; 来自腾讯; 人工标注的数据集CSCD-NS;
10.mcsc_tet.5000.json(5000): 医学拼写纠错; 来自腾讯医典APP的真实历史日志; 注意论文说该数据集只关注医学实体的纠错, 常用字等的纠错并不关注;
11.ecspell.dev.json(1500): 来自ECSpell论文, 包括(law/med/gov)等三个领域;
12.sighan2013.dev.json(1000): 来自sighan13会议;
13.sighan2014.dev.json(1062): 来自sighan14会议;
14.sighan2015.dev.json(1100): 来自sighan15会议;
1.2 测评数据预处理
测评数据都经过 全角转半角,繁简转化,标点符号标准化等操作;
1.3 其他说明
1.指标带common的极为宽松指标, 同开源项目pycorrector的评估指标;
2.指标带strict的极为严格指标, 同开源项目[wangwang110/CSC](https://github.com/wangwang110/CSC);
3.macbert4mdcspell_v1模型为训练使用mdcspell架构+bert的mlm-loss, 但是推理的时候只用bert-mlm;
4.acc_rmrb/acc_xxqg数据集没有错误, 用于评估模型的误纠率(过度纠错);
5.qwen25_1-5b_pycorrector的模型为shibing624/chinese-text-correction-1.5b, 其训练数据包括了lemon_v2/mcsc_tet/ecspell的验证集和测试集, 其他的bert类模型的训练不包括验证集和测试集;
二、重要指标
2.1 F1(common_cor_f1)
model/common_cor_f1 | avg | gen_de3 | lemon_v2 | gen_passage | text_proof | gen_xxqg | faspell | lomo_tet | mcsc_tet | ecspell | sighan2013 | sighan2014 | sighan2015 |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
macbert4csc_pycorrector | 45.8 | 42.44 | 42.89 | 31.49 | 46.31 | 26.06 | 32.7 | 44.83 | 27.93 | 55.51 | 70.89 | 61.72 | 66.81 |
bert4csc_v1 | 62.28 | 93.73 | 61.99 | 44.79 | 68.0 | 35.03 | 48.28 | 61.8 | 64.41 | 79.11 | 77.66 | 51.01 | 61.54 |
macbert4csc_v1 | 68.55 | 96.67 | 65.63 | 48.4 | 75.65 | 38.43 | 51.76 | 70.11 | 80.63 | 85.55 | 81.38 | 57.63 | 70.7 |
macbert4csc_v2 | 68.6 | 96.74 | 66.02 | 48.26 | 75.78 | 38.84 | 51.91 | 70.17 | 80.71 | 85.61 | 80.97 | 58.22 | 69.95 |
macbert4mdcspell_v1 | 71.1 | 96.42 | 70.06 | 52.55 | 79.61 | 43.37 | 53.85 | 70.9 | 82.38 | 87.46 | 84.2 | 61.08 | 71.32 |
qwen25_1-5b_pycorrector | 45.11 | 27.29 | 89.48 | 14.61 | 83.9 | 13.84 | 18.2 | 36.71 | 96.29 | 88.2 | 36.41 | 15.64 | 20.73 |
2.2 acc(common_cor_acc)
model/common_cor_acc | avg | gen_de3 | lemon_v2 | gen_passage | text_proof | gen_xxqg | faspell | lomo_tet | mcsc_tet | ecspell | sighan2013 | sighan2014 | sighan2015 |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
macbert4csc_pycorrector | 48.26 | 26.96 | 28.68 | 34.16 | 55.29 | 28.38 | 22.2 | 60.96 | 57.16 | 67.73 | 55.9 | 68.93 | 72.73 |
bert4csc_v1 | 60.76 | 88.21 | 45.96 | 43.13 | 68.97 | 35.0 | 34.0 | 65.86 | 73.26 | 81.8 | 64.5 | 61.11 | 67.27 |
macbert4csc_v1 | 65.34 | 93.56 | 49.76 | 44.98 | 74.64 | 36.1 | 37.0 | 73.0 | 83.6 | 86.87 | 69.2 | 62.62 | 72.73 |
macbert4csc_v2 | 65.22 | 93.69 | 50.14 | 44.92 | 74.64 | 36.26 | 37.0 | 72.72 | 83.66 | 86.93 | 68.5 | 62.43 | 71.73 |
macbert4mdcspell_v1 | 67.15 | 93.09 | 54.8 | 47.71 | 78.09 | 39.52 | 38.8 | 71.92 | 84.78 | 88.27 | 73.2 | 63.28 | 72.36 |
qwen25_1-5b_pycorrector | 46.09 | 15.82 | 81.29 | 22.96 | 82.17 | 19.04 | 12.8 | 50.2 | 96.4 | 89.13 | 22.8 | 27.87 | 32.55 |
2.3 acc(acc_true, thr=0.75)
model/acc | avg | acc_rmrb | acc_xxqg |
---|---|---|---|
macbert4csc_pycorrector | 99.24 | 99.22 | 99.26 |
bert4csc_v1 | 98.71 | 98.36 | 99.06 |
macbert4csc_v1 | 97.72 | 96.72 | 98.72 |
macbert4csc_v2 | 97.89 | 96.98 | 98.8 |
macbert4mdcspell_v1 | 97.75 | 96.51 | 98.98 |
qwen25_1-5b_pycorrector | 82.0 | 77.14 | 86.86 |
二、结论(Conclusion)
1.macbert4csc_v1/macbert4csc_v2/macbert4mdcspell_v1等模型使用多种领域数据训练, 比较均衡, 也适合作为第一步的预训练模型, 可用于专有领域数据的继续微调;
2.比较macbert4csc_pycorrector/bertbase4csc_v1/macbert4csc_v2/macbert4mdcspell_v1, 观察表2.3, 可以发现训练数据越多, 准确率提升的同时, 误纠率也会稍微高一些;
3.MFT(Mask-Correct)依旧有效, 不过对于数据量足够的情形提升不明显, 可能也是误纠率升高的一个重要原因;
4.训练数据中也存在文言文数据, 训练好的模型也支持文言文纠错;
5.训练好的模型对"地得的"等高频错误具有较高的识别率和纠错率;
三、调用(Usage)
3.1 使用macro-correct
import os
os.environ["MACRO_CORRECT_FLAG_CSC_TOKEN"] = "1"
from macro_correct import correct
### 默认纠错(list输入)
text_list = ["真麻烦你了。希望你们好好的跳无",
"少先队员因该为老人让坐",
"机七学习是人工智能领遇最能体现智能的一个分知",
"一只小鱼船浮在平净的河面上"
]
text_csc = correct(text_list)
print("默认纠错(list输入):")
for res_i in text_csc:
print(res_i)
print("#" * 128)
"""
默认纠错(list输入):
{'index': 0, 'source': '真麻烦你了。希望你们好好的跳无', 'target': '真麻烦你了。希望你们好好地跳舞', 'errors': [['的', '地', 12, 0.6584], ['无', '舞', 14, 1.0]]}
{'index': 1, 'source': '少先队员因该为老人让坐', 'target': '少先队员应该为老人让坐', 'errors': [['因', '应', 4, 0.995]]}
{'index': 2, 'source': '机七学习是人工智能领遇最能体现智能的一个分知', 'target': '机器学习是人工智能领域最能体现智能的一个分支', 'errors': [['七', '器', 1, 0.9998], ['遇', '域', 10, 0.9999], ['知', '支', 21, 1.0]]}
{'index': 3, 'source': '一只小鱼船浮在平净的河面上', 'target': '一只小鱼船浮在平静的河面上', 'errors': [['净', '静', 8, 0.9961]]}
"""
3.2 使用 transformers
# !/usr/bin/python
# -*- coding: utf-8 -*-
# @time : 2021/2/29 21:41
# @author : Mo
# @function: transformers直接加载bert类模型测试
import traceback
import time
import sys
import os
os.environ["USE_TORCH"] = "1"
from transformers import BertConfig, BertTokenizer, BertForMaskedLM
import torch
# pretrained_model_name_or_path = "shibing624/macbert4csc-base-chinese"
# pretrained_model_name_or_path = "Macropodus/macbert4mdcspell_v1"
# pretrained_model_name_or_path = "Macropodus/macbert4csc_v1"
pretrained_model_name_or_path = "Macropodus/macbert4csc_v2"
# pretrained_model_name_or_path = "Macropodus/bert4csc_v1"
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
max_len = 128
print("load model, please wait a few minute!")
tokenizer = BertTokenizer.from_pretrained(pretrained_model_name_or_path)
bert_config = BertConfig.from_pretrained(pretrained_model_name_or_path)
model = BertForMaskedLM.from_pretrained(pretrained_model_name_or_path)
model.to(device)
print("load model success!")
texts = [
"机七学习是人工智能领遇最能体现智能的一个分知",
"我是练习时长两念半的鸽仁练习生蔡徐坤",
"真麻烦你了。希望你们好好的跳无",
"他法语说的很好,的语也不错",
"遇到一位很棒的奴生跟我疗天",
"我们为这个目标努力不解",
]
len_mid = min(max_len, max([len(t)+2 for t in texts]))
with torch.no_grad():
outputs = model(**tokenizer(texts, padding=True, max_length=len_mid,
return_tensors="pt").to(device))
def get_errors(source, target):
""" 极简方法获取 errors """
len_min = min(len(source), len(target))
errors = []
for idx in range(len_min):
if source[idx] != target[idx]:
errors.append([source[idx], target[idx], idx])
return errors
result = []
for probs, source in zip(outputs.logits, texts):
ids = torch.argmax(probs, dim=-1)
tokens_space = tokenizer.decode(ids[1:-1], skip_special_tokens=False)
text_new = tokens_space.replace(" ", "")
target = text_new[:len(source)]
errors = get_errors(source, target)
print(source, " => ", target, errors)
result.append([target, errors])
print(result)
"""
机七学习是人工智能领遇最能体现智能的一个分知 => 机器学习是人工智能领域最能体现智能的一个分支 [['七', '器', 1], ['遇', '域', 10], ['知', '支', 21]]
我是练习时长两念半的鸽仁练习生蔡徐坤 => 我是练习时长两年半的个人练习生蔡徐坤 [['念', '年', 7], ['鸽', '个', 10], ['仁', '人', 11]]
真麻烦你了。希望你们好好的跳无 => 真麻烦你了。希望你们好好地跳舞 [['的', '地', 12], ['无', '舞', 14]]
他法语说的很好,的语也不错 => 他法语说得很好,德语也不错 [['的', '得', 4], ['的', '德', 8]]
遇到一位很棒的奴生跟我疗天 => 遇到一位很棒的女生跟我聊天 [['奴', '女', 7], ['疗', '聊', 11]]
我们为这个目标努力不解 => 我们为这个目标努力不懈 [['解', '懈', 10]]
"""
四、论文(Paper)
2024-Refining: Refining Corpora from a Model Calibration Perspective for Chinese
2024-ReLM: Chinese Spelling Correction as Rephrasing Language Model
2024-DICS: DISC: Plug-and-Play Decoding Intervention with Similarity of Characters for Chinese Spelling Check
2023-Bi-DCSpell: A Bi-directional Detector-Corrector Interactive Framework for Chinese Spelling Check
2023-BERT-MFT: Rethinking Masked Language Modeling for Chinese Spelling Correction
2023-PTCSpell: PTCSpell: Pre-trained Corrector Based on Character Shape and Pinyin for Chinese Spelling Correction
2023-DR-CSC: A Frustratingly Easy Plug-and-Play Detection-and-Reasoning Module for Chinese
2023-DROM: Disentangled Phonetic Representation for Chinese Spelling Correction
2023-EGCM: An Error-Guided Correction Model for Chinese Spelling Error Correction
2023-IGPI: Investigating Glyph-Phonetic Information for Chinese Spell Checking: What Works and What’s Next?
2022-CRASpell: CRASpell: A Contextual Typo Robust Approach to Improve Chinese Spelling Correction
2022-MDCSpell: MDCSpell: A Multi-task Detector-Corrector Framework for Chinese Spelling Correction
2021-MLMPhonetics: Correcting Chinese Spelling Errors with Phonetic Pre-training
2021-ChineseBERT: ChineseBERT: Chinese Pretraining Enhanced by Glyph and Pinyin Information
2021-BERTCrsGad: Global Attention Decoder for Chinese Spelling Error Correction
2021-ThinkTwice: Think Twice: A Post-Processing Approach for the Chinese Spelling Error Correction
2021-PHMOSpell: PHMOSpell: Phonological and Morphological Knowledge Guided Chinese Spelling Chec
2021-SpellBERT: SpellBERT: A Lightweight Pretrained Model for Chinese Spelling Check
2021-TwoWays: Exploration and Exploitation: Two Ways to Improve Chinese Spelling Correction Models
2021-ReaLiSe: Read, Listen, and See: Leveraging Multimodal Information Helps Chinese Spell Checking
2021-DCSpell: DCSpell: A Detector-Corrector Framework for Chinese Spelling Error Correction
2021-PLOME: PLOME: Pre-training with Misspelled Knowledge for Chinese Spelling Correction
2021-DCN: Dynamic Connected Networks for Chinese Spelling Check
2020-SoftMaskBERT: Spelling Error Correction with Soft-Masked BERT
2020-SpellGCN: SpellGCN:Incorporating Phonological and Visual Similarities into Language Models for Chinese Spelling Check
2020-ChunkCSC: Chunk-based Chinese Spelling Check with Global Optimization
2020-MacBERT: Revisiting Pre-Trained Models for Chinese Natural Language Processing
2019-FASPell: FASPell: A Fast, Adaptable, Simple, Powerful Chinese Spell Checker Based On DAE-Decoder Paradigm
2018-Hybrid: A Hybrid Approach to Automatic Corpus Generation for Chinese Spelling Checking
2015-Sighan15: Introduction to SIGHAN 2015 Bake-off for Chinese Spelling Check
2014-Sighan14: Overview of SIGHAN 2014 Bake-off for Chinese Spelling Check
2013-Sighan13: Chinese Spelling Check Evaluation at SIGHAN Bake-off 2013
五、参考(Refer)
- nghuyong/Chinese-text-correction-papers
- destwang/CTCResources
- wangwang110/CSC
- chinese-poetry/chinese-poetry
- chinese-poetry/huajianji
- garychowcmu/daizhigev20
- yangjianxin1/Firefly
- Macropodus/xuexiqiangguo_428w
- Macropodus/csc_clean_wang271k
- Macropodus/csc_eval_public
- shibing624/pycorrector
- iioSnail/MDCSpell_pytorch
- gingasan/lemon
- Claude-Liu/ReLM
六、引用(Cite)
For citing this work, you can refer to the present GitHub project. For example, with BibTeX:
@software{macro-correct,
url = {https://github.com/yongzhuo/macro-correct},
author = {Yongzhuo Mo},
title = {macro-correct},
year = {2025}
- Downloads last month
- 1
Model tree for Macropodus/macbert4csc_v2
Base model
hfl/chinese-macbert-base