PyTorch impelementations of BERT-based Spelling Error Correction Models.

Overview

BertBasedCorrectionModels

基于BERT的文本纠错模型,使用PyTorch实现

数据准备

  1. http://nlp.ee.ncu.edu.tw/resource/csc.html下载SIGHAN数据集
  2. 解压上述数据集并将文件夹中所有 ''.sgml'' 文件复制至 datasets/csc/ 目录
  3. 复制 ''SIGHAN15_CSC_TestInput.txt'' 和 ''SIGHAN15_CSC_TestTruth.txt'' 至 datasets/csc/ 目录
  4. 下载 https://github.com/wdimmy/Automatic-Corpus-Generation/blob/master/corpus/train.sgml 至 datasets/csc 目录
  5. 请确保以下文件在 datasets/csc 中
    train.sgml
    B1_training.sgml
    C1_training.sgml  
    SIGHAN15_CSC_A2_Training.sgml  
    SIGHAN15_CSC_B2_Training.sgml  
    SIGHAN15_CSC_TestInput.txt
    SIGHAN15_CSC_TestTruth.txt
    

环境准备

  1. 使用已有编码环境或通过 conda create -n python=3.7 创建一个新环境(推荐)
  2. 克隆本项目并进入项目根目录
  3. 安装所需依赖 pip install -r requirements.txt
  4. 如果出现报错 GLIBC 版本过低的问题(GLIBC 的版本更迭容易出事故,不推荐更新),openCC 改为安装较低版本(例如 1.1.0)
  5. 在当前终端将此目录加入环境变量 export PYTHONPATH=.

训练

运行以下命令以训练模型,首次运行会自动处理数据。

python tools/train_csc.py --config_file csc/train_SoftMaskedBert.yml

可选择不同配置文件以训练不同模型,目前支持以下配置文件:

  • train_bert4csc.yml
  • train_macbert4csc.yml
  • train_SoftMaskedBert.yml

如有其他需求,可根据需要自行调整配置文件中的参数。

实验结果

SoftMaskedBert

component sentence level acc p r f
Detection 0.5045 0.8252 0.8416 0.8333
Correction 0.8055 0.9395 0.8748 0.9060

Bert类

char level

MODEL p r f
BERT4CSC 0.9269 0.8651 0.8949
MACBERT4CSC 0.9380 0.8736 0.9047

sentence level

model acc p r f
BERT4CSC 0.7990 0.8482 0.7214 0.7797
MACBERT4CSC 0.8027 0.8525 0.7251 0.7836

推理

方法一,使用inference脚本:

python inference.py --ckpt_fn epoch=0-val_loss=0.03.ckpt --texts "我今天很高心"
# 或给出line by line格式的文本地址
python inference.py --ckpt_fn epoch=0-val_loss=0.03.ckpt --text_file /ml/data/text.txt

其中/ml/data/text.txt文本如下:

我今天很高心
你这个辣鸡模型只能做错别字纠正

方法二,直接调用

texts = ['今天我很高心', '测试', '继续测试']
model.predict(texts)

方法三、导出bert权重,使用transformers或pycorrector调用

  1. 使用convert_to_pure_state_dict.py导出bert权重
  2. 后续步骤参考https://github.com/shibing624/pycorrector/blob/master/pycorrector/macbert/README.md

引用

如果你在研究中使用了本项目,请按如下格式引用:

@article{cai2020pre,
  title={BERT Based Correction Models},
  author={Cai, Heng and Chen, Dian},
  journal={GitHub. Note: https://github.com/gitabtion/BertBasedCorrectionModels},
  year={2020}
}

License

本源代码的授权协议为 Apache License 2.0,可免费用做商业用途。请在产品说明中附加本项目的链接和授权协议。本项目受版权法保护,侵权必究。

更新记录

20210618

  1. 修复数据处理的编码报错问题

20210518

  1. 将BERT4CSC检错任务改为使用FocalLoss
  2. 更新修改后的模型实验结果
  3. 降低数据处理时保留原文的概率

20210517

  1. 对BERT4CSC模型新增检错任务
  2. 新增基于LineByLine文件的inference

References

  1. Spelling Error Correction with Soft-Masked BERT
  2. http://ir.itc.ntnu.edu.tw/lre/sighan8csc.html
  3. https://github.com/wdimmy/Automatic-Corpus-Generation
  4. transformers
  5. https://github.com/sunnyqiny/Confusionset-guided-Pointer-Networks-for-Chinese-Spelling-Check
  6. SoftMaskedBert-PyTorch
  7. Deep-Learning-Project-Template
  8. https://github.com/lonePatient/TorchBlocks
  9. https://github.com/shibing624/pycorrector
Owner
Heng Cai
NLPer
Heng Cai
PyTorch code for EMNLP 2019 paper "LXMERT: Learning Cross-Modality Encoder Representations from Transformers".

LXMERT: Learning Cross-Modality Encoder Representations from Transformers Our servers break again :(. I have updated the links so that they should wor

Hao Tan 838 Dec 19, 2022
Repository to hold code for the cap-bot varient that is being presented at the SIIC Defence Hackathon 2021.

capbot-siic Repository to hold code for the cap-bot varient that is being presented at the SIIC Defence Hackathon 2021. Problem Inspiration A plethora

Aryan Kargwal 19 Feb 17, 2022
This repository contains (not all) code from my project on Named Entity Recognition in philosophical text

NERphilosophy 👋 Welcome to the github repository of my BsC thesis. This repository contains (not all) code from my project on Named Entity Recognitio

Ruben 1 Jan 27, 2022
Russian GPT3 models.

Russian GPT-3 models (ruGPT3XL, ruGPT3Large, ruGPT3Medium, ruGPT3Small) trained with 2048 sequence length with sparse and dense attention blocks. We also provide Russian GPT-2 large model (ruGPT2Larg

Sberbank AI 1.6k Jan 05, 2023
NLP topic mdel LDA - Gathered from New York Times website

NLP topic mdel LDA - Gathered from New York Times website

1 Oct 14, 2021
sangha, pronounced "suhng-guh", is a social networking, booking platform where students and teachers can share their practice.

Flask React Project This is the backend for the Flask React project. Getting started Clone this repository (only this branch) git clone https://github

Courtney Newcomer 17 Sep 29, 2021
Model for recasing and repunctuating ASR transcripts

Recasing and punctuation model based on Bert Benoit Favre 2021 This system converts a sequence of lowercase tokens without punctuation to a sequence o

Benoit Favre 88 Dec 29, 2022
SimCSE: Simple Contrastive Learning of Sentence Embeddings

SimCSE: Simple Contrastive Learning of Sentence Embeddings This repository contains the code and pre-trained models for our paper SimCSE: Simple Contr

Princeton Natural Language Processing 2.5k Jan 07, 2023
Behavioral Testing of Clinical NLP Models

Behavioral Testing of Clinical NLP Models This repository contains code for testing the behavior of clinical prediction models based on patient letter

Betty van Aken 2 Sep 20, 2022
Big Bird: Transformers for Longer Sequences

BigBird, is a sparse-attention based transformer which extends Transformer based models, such as BERT to much longer sequences. Moreover, BigBird comes along with a theoretical understanding of the c

Google Research 457 Dec 23, 2022
precise iris segmentation

PI-DECODER Introduction PI-DECODER, a decoder structure designed for Precise Iris Segmentation and Location. The decoder structure is shown below: Ple

8 Aug 08, 2022
NLP and Text Generation Experiments in TensorFlow 2.x / 1.x

Code has been run on Google Colab, thanks Google for providing computational resources Contents Natural Language Processing(自然语言处理) Text Classificati

1.5k Nov 14, 2022
基于Transformer的单模型、多尺度的VAE模型

UniVAE 基于Transformer的单模型、多尺度的VAE模型 介绍 https://kexue.fm/archives/8475 依赖 需要大于0.10.6版本的bert4keras(当前还没有推到pypi上,可以直接从GitHub上clone最新版)。 引用 @misc{univae,

苏剑林(Jianlin Su) 49 Aug 24, 2022
Code for text augmentation method leveraging large-scale language models

HyperMix Code for our paper GPT3Mix and conducting classification experiments using GPT-3 prompt-based data augmentation. Getting Started Installing P

NAVER AI 47 Dec 20, 2022
Persian-lexicon - A lexicon of 70K unique Persian (Farsi) words

Persian Lexicon This repo uses Uppsala Persian Corpus (UPC) to construct a lexic

Saman Vaisipour 7 Apr 01, 2022
Wrapper to display a script output or a text file content on the desktop in sway or other wlroots-based compositors

nwg-wrapper This program is a part of the nwg-shell project. This program is a GTK3-based wrapper to display a script output, or a text file content o

Piotr Miller 94 Dec 27, 2022
Contact Extraction with Question Answering.

contactsQA Extraction of contact entities from address blocks and imprints with Extractive Question Answering. Goal Input: Dr. Max Mustermann Hauptstr

Jan 2 Apr 20, 2022
Quick insights from Zoom meeting transcripts using Graph + NLP

Transcript Analysis - Graph + NLP This program extracts insights from Zoom Meeting Transcripts (.vtt) using TigerGraph and NLTK. In order to run this

Advit Deepak 7 Sep 17, 2022
This repository contains the code, models and datasets discussed in our paper "Few-Shot Question Answering by Pretraining Span Selection"

Splinter This repository contains the code, models and datasets discussed in our paper "Few-Shot Question Answering by Pretraining Span Selection", to

Ori Ram 88 Dec 31, 2022
NVDA, the free and open source Screen Reader for Microsoft Windows

NVDA NVDA (NonVisual Desktop Access) is a free, open source screen reader for Microsoft Windows. It is developed by NV Access in collaboration with a

NV Access 1.6k Jan 07, 2023