Chinese named entity recognization (bert/roberta/macbert/bert_wwm with Keras)

Related tags

Text Data & NLPBERT-NER-Keras
Overview

Chinese named entity recognization (bert/roberta/macbert/bert_wwm with Keras)

Project Structure

./
├── DataProcess
│   ├── __pycache__
│   ├── convert2bio.py
│   ├── convert_jsonl.py
│   ├── handle_numbers.py
│   ├── load_data.py
│   └── statistic.py
├── README.md
├── __pycache__
├── chinese_L-12_H-768_A-12                                    BERT权重
│   ├── bert_config.json
│   ├── bert_model.ckpt.data-00000-of-00001
│   ├── bert_model.ckpt.index
│   ├── bert_model.ckpt.meta
│   └── vocab.txt
├── chinese_bert_wwm                                           BERT_wwm权重
│   ├── bert_config.json
│   ├── bert_model.ckpt.data-00000-of-00001
│   ├── bert_model.ckpt.index
│   ├── bert_model.ckpt.meta
│   └── vocab.txt
├── chinese_macbert_base                                       macBERT权重
│   ├── chinese_macbert_base.ckpt.data-00000-of-00001
│   ├── chinese_macbert_base.ckpt.index
│   ├── chinese_macbert_base.ckpt.meta
│   ├── macbert_base_config.json
│   └── vocab.txt
├── chinese_roberta_wwm_ext_L-12_H-768_A-12                    roberta权重
│   ├── bert_config.json
│   ├── bert_model.ckpt.data-00000-of-00001
│   ├── bert_model.ckpt.index
│   ├── bert_model.ckpt.meta
│   └── vocab.txt
├── config                                                     
│   ├── __pycache__
│   ├── config.py                                              配置文件
│   └── pulmonary_label2id.json                                label id
├── data                                                       数据集
│   ├── pulmonary.test
│   ├── pulmonary.train
│   └── sict_train.txt
├── environment.yaml                                           conda环境配置文件
├── evaluate.py
├── generator_train.py
├── keras_bert                                                 keras_bert（可pip下）
├── keras_contrib                                              keras_contrib（可pip下）
├── log                                                        训练nohup日志
│   ├── chinese_L-12_H-768_A-12.out
│   ├── chinese_macbert_base.out
│   ├── chinese_roberta_wwm_ext_L-12_H-768_A-12.out
│   └── electra_180g_base.out
├── model.py                                                   模型构建文件
├── models                                                     保存的模型权重
│   ├── pulmonary_chinese_L-12_H-768_A-12_ner.h5
│   ├── pulmonary_chinese_bert_wwm_ner.h5
│   ├── pulmonary_chinese_macbert_base_ner.h5
│   └── pulmonary_chinese_roberta_wwm_ext_L-12_H-768_A-12_ner.h5
├── predict.py                                                 预测
├── report                                                     模型实体F1评估报告
│   ├── pulmonary_chinese_L-12_H-768_A-12_evaluate.txt
│   ├── pulmonary_chinese_L-12_H-768_A-12_predict.json
│   ├── pulmonary_chinese_bert_wwm_evaluate.txt
│   ├── pulmonary_chinese_bert_wwm_predict.json
│   ├── pulmonary_chinese_macbert_base_evaluate.txt
│   ├── pulmonary_chinese_macbert_base_predict.json
│   ├── pulmonary_chinese_roberta_wwm_ext_L-12_H-768_A-12_evaluate.txt
│   └── pulmonary_chinese_roberta_wwm_ext_L-12_H-768_A-12_predict.json
├── requirements.txt                                           pip环境
├── test.py                                                    
├── train.py                                                   训练
└── utils                                                      
    ├── FGM.py                                                 FGM对抗
    ├── __pycache__
    └── path.py                                                所有路径

56 directories, 193 files

Dataset

三甲医院肺结节数据集，20000+字，BIO格式，形如：

中	B-ORG
共	I-ORG
中	I-ORG
央	I-ORG
致	O
中	B-ORG
国	I-ORG
致	I-ORG
公	I-ORG
党	I-ORG
十	I-ORG
一	I-ORG
大	I-ORG
的	O
贺	O
词	O

ATTENTION: 在处理自己数据集的时候需要注意：

  • 字与标签之间用空格（"\ "）隔开
  • 其中句子与句子之间使用空行隔开

Steps

  1. 替换数据集
  2. 使用DataProcess/load_data.py生成label2id.txt文件
  3. 修改config/config.py中的MAX_SEQ_LEN（超过截断，少于填充，最好设置训练集、测试集中最长句子作为MAX_SEQ_LEN）
  4. 下载权重，放到项目中
  5. 修改public/path.py中的地址
  6. 根据需要修改model.py模型结构
  7. 修改config/config.py的参数
  8. 训练前debug看下input_train_labels,result_train对不对，input_train_types全是0
  9. 训练

Model

BERT

roberta

macBERT

BERT_wwm

Train

运行train.py

Evaluate

运行evaluate/f1_score.py

BERT

           precision    recall  f1-score   support

     SIGN     0.6651    0.7354    0.6985       189
  ANATOMY     0.8333    0.8409    0.8371       220
 DIAMETER     1.0000    1.0000    1.0000        16
  DISEASE     0.4915    0.6744    0.5686        43
 QUANTITY     0.8837    0.9157    0.8994        83
TREATMENT     0.3571    0.5556    0.4348         9
  DENSITY     1.0000    1.0000    1.0000         8
    ORGAN     0.4500    0.6923    0.5455        13
LUNGFIELD     1.0000    0.5000    0.6667         6
    SHAPE     0.5714    0.5714    0.5714         7
   NATURE     1.0000    1.0000    1.0000         6
 BOUNDARY     1.0000    0.6250    0.7692         8
   MARGIN     0.8333    0.8333    0.8333         6
  TEXTURE     1.0000    0.8571    0.9231         7

micro avg     0.7436    0.7987    0.7702       621
macro avg     0.7610    0.7987    0.7760       621

roberta

           precision    recall  f1-score   support

  ANATOMY     0.8624    0.8545    0.8584       220
  DENSITY     0.8000    1.0000    0.8889         8
     SIGN     0.7347    0.7619    0.7481       189
 QUANTITY     0.8977    0.9518    0.9240        83
  DISEASE     0.5690    0.7674    0.6535        43
 DIAMETER     1.0000    1.0000    1.0000        16
TREATMENT     0.3333    0.5556    0.4167         9
 BOUNDARY     1.0000    0.6250    0.7692         8
LUNGFIELD     1.0000    0.6667    0.8000         6
   MARGIN     0.8333    0.8333    0.8333         6
  TEXTURE     1.0000    0.8571    0.9231         7
    SHAPE     0.5714    0.5714    0.5714         7
   NATURE     1.0000    1.0000    1.0000         6
    ORGAN     0.6250    0.7692    0.6897        13

micro avg     0.7880    0.8261    0.8066       621
macro avg     0.8005    0.8261    0.8104       621

macBERT

           precision    recall  f1-score   support

  ANATOMY     0.8773    0.8773    0.8773       220
     SIGN     0.6538    0.7196    0.6851       189
  DISEASE     0.5893    0.7674    0.6667        43
 QUANTITY     0.9070    0.9398    0.9231        83
    ORGAN     0.5882    0.7692    0.6667        13
  TEXTURE     1.0000    0.8571    0.9231         7
 DIAMETER     1.0000    1.0000    1.0000        16
TREATMENT     0.3750    0.6667    0.4800         9
LUNGFIELD     1.0000    0.5000    0.6667         6
    SHAPE     0.4286    0.4286    0.4286         7
   NATURE     1.0000    1.0000    1.0000         6
  DENSITY     1.0000    1.0000    1.0000         8
 BOUNDARY     1.0000    0.6250    0.7692         8
   MARGIN     0.8333    0.8333    0.8333         6

micro avg     0.7697    0.8180    0.7931       621
macro avg     0.7846    0.8180    0.7977       621

BERT_wwm

           precision    recall  f1-score   support

  DISEASE     0.5667    0.7907    0.6602        43
  ANATOMY     0.8676    0.8636    0.8656       220
 QUANTITY     0.8966    0.9398    0.9176        83
     SIGN     0.7358    0.7513    0.7435       189
LUNGFIELD     1.0000    0.6667    0.8000         6
TREATMENT     0.3571    0.5556    0.4348         9
 DIAMETER     0.9375    0.9375    0.9375        16
 BOUNDARY     1.0000    0.6250    0.7692         8
  TEXTURE     1.0000    0.8571    0.9231         7
   MARGIN     0.8333    0.8333    0.8333         6
    ORGAN     0.5882    0.7692    0.6667        13
  DENSITY     1.0000    1.0000    1.0000         8
   NATURE     1.0000    1.0000    1.0000         6
    SHAPE     0.5000    0.5714    0.5333         7

micro avg     0.7889    0.8245    0.8063       621
macro avg     0.8020    0.8245    0.8104       621

Predict

运行predict/predict_bio.py

Owner
GitHub Repository
Auto_code_complete is a auto word-completetion program which allows you to customize it on your needs

auto_code_complete is a auto word-completetion program which allows you to customize it on your needs. the model for this program is one of the deep-learning NLP(Natural Language Process) model struc

RUO 2 Feb 22, 2022
🏆 • 5050 most frequent words in 109 languages

🏆 Most Common Words Multilingual 5000 most frequent words in 109 languages. Uses wordfrequency.info as a source. 🔗 License source code license data

14 Nov 24, 2022
Textpipe: clean and extract metadata from text

textpipe: clean and extract metadata from text textpipe is a Python package for converting raw text in to clean, readable text and extracting metadata

Textpipe 298 Nov 21, 2022
A Python module made to simplify the usage of Text To Speech and Speech Recognition.

Nav Module The solution for voice related stuff in Python Nav is a Python module which simplifies voice related stuff in Python. Just import the Modul

Snm Logic 1 Dec 20, 2021
Proquabet - Convert your prose into proquints and then you essentially have Vogon poetry

Proquabet Turn your prose into a constant stream of encrypted and meaningless-so

Milo Fultz 2 Oct 10, 2022
A tool helps build a talk preview image by combining the given background image and talk event description

talk-preview-img-builder A tool helps build a talk preview image by combining the given background image and talk event description Installation and U

PyCon Taiwan 4 Aug 20, 2022
A Fast Sequence Transducer Implementation with PyTorch Bindings

transducer A Fast Sequence Transducer Implementation with PyTorch Bindings. The corresponding publication is Sequence Transduction with Recurrent Neur

Awni Hannun 184 Dec 18, 2022
HiFi DeepVariant + WhatsHap workflowHiFi DeepVariant + WhatsHap workflow

HiFi DeepVariant + WhatsHap workflow Workflow steps align HiFi reads to reference with pbmm2 call small variants with DeepVariant, using two-pass meth

William Rowell 2 May 14, 2022
Fast, DB Backed pretrained word embeddings for natural language processing.

Embeddings Embeddings is a python package that provides pretrained word embeddings for natural language processing and machine learning. Instead of lo

Victor Zhong 212 Nov 21, 2022
Simple Python script to scrape youtube channles of "Parity Technologies and Web3 Foundation" and translate them to well-known braille language or any language

Simple Python script to scrape youtube channles of "Parity Technologies and Web3 Foundation" and translate them to well-known braille language or any

Little Endian 1 Apr 28, 2022
An open source framework for seq2seq models in PyTorch.

pytorch-seq2seq Documentation This is a framework for sequence-to-sequence (seq2seq) models implemented in PyTorch. The framework has modularized and

International Business Machines 1.4k Jan 02, 2023
Plugin repository for Macast

Macast-plugins Plugin repository for Macast. How to use third-party player plugin Download Macast from GitHub Release. Download the plugin you want fr

109 Jan 04, 2023
Türkçe küfürlü içerikleri bulan bir yapay zeka kütüphanesi / An ML library for profanity detection in Turkish sentences

"Kötü söz sahibine aittir." -Anonim Nedir? sinkaf uygunsuz yorumların bulunmasını sağlayan bir python kütüphanesidir. Farkı nedir? Diğer algoritmalard

KaraGoz 4 Feb 18, 2022
BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese

Table of contents Introduction Using BARTpho with fairseq Using BARTpho with transformers Notes BARTpho: Pre-trained Sequence-to-Sequence Models for V

VinAI Research 58 Dec 23, 2022
translate using your voice

speech-to-text-translator Usage translate using your voice description this project makes translating a word easy, all you have to do is speak and...

1 Oct 18, 2021
This is the library for the Unbounded Interleaved-State Recurrent Neural Network (UIS-RNN) algorithm, corresponding to the paper Fully Supervised Speaker Diarization.

UIS-RNN Overview This is the library for the Unbounded Interleaved-State Recurrent Neural Network (UIS-RNN) algorithm. UIS-RNN solves the problem of s

Google 1.4k Dec 28, 2022
Chatbot with Pytorch, Python & Nextjs

Installation Instructions Make sure that you have Python 3, gcc, venv, and pip installed. Clone the repository $ git clone https://github.com/sahr

Rohit Sah 0 Dec 11, 2022
An Analysis Toolkit for Natural Language Generation (Translation, Captioning, Summarization, etc.)

VizSeq is a Python toolkit for visual analysis on text generation tasks like machine translation, summarization, image captioning, speech translation

Facebook Research 409 Oct 28, 2022
Simple translation demo showcasing our headliner package.

Headliner Demo This is a demo showcasing our Headliner package. In particular, we trained a simple seq2seq model on an English-German dataset. We didn

Axel Springer News Media & Tech GmbH & Co. KG - Ideas Engineering 16 Nov 24, 2022
The NewSHead dataset is a multi-doc headline dataset used in NHNet for training a headline summarization model.

This repository contains the raw dataset used in NHNet [1] for the task of News Story Headline Generation. The code of data processing and training is available under Tensorflow Models - NHNet.

Google Research Datasets 31 Jul 15, 2022