Skip to content

A simple BiLSTM-CRF model for Chinese Named Entity Recognition 中文命名实体识别 (TensorFlow)

Notifications You must be signed in to change notification settings

ninghongbo123/zh-NER-TF

 
 

Repository files navigation

A simple BiLSTM-CRF model for Chinese Named Entity Recognition

This repository includes the code for buliding a very simple character-based BiLSTM-CRF sequence labelling model for Chinese Named Entity Recognition task. Its goal is to recognize three types of Named Entity: PERSON, LOCATION and ORGANIZATION.

This code works on Python 3 & TensorFlow 1.2 and the following repository https://github.com/guillaumegenthial/sequence_tagging gives me much help.

model

This model is similar to the models provied by paper [1] and [2]. Its structure looks just like the following illustration:

Network

For one Chinese sentence, each character in this sentence has / will have a tag which belongs to the set {O, B-PER, I-PER, B-LOC, I-LOC, B-ORG, I-ORG}.

The first layer, look-up layer, aims at transforming character representation from one-hot vector into character embedding. In this code I initialize the embedding matrix randomly and I know it looks too simple. We could add some language knowledge later. For example, do tokenization and use pre-trained word-level embedding, then every character in one token could be initialized with this token's word embedding. In addition, we can get the character embedding by combining low-level features (please see paper[2]'s section 4.1 and paper[3]'s section 3.3 for more details).

The second layer, BiLSTM layer, can efficiently use both past and future input information and extract features automatically.

The third layer, CRF layer, labels the tag for each character in one sentence. If we use Softmax layer for labelling we might get ungrammatic tag sequences beacuse Softmax could only label each position independently. We know that 'I-LOC' cannot follow 'B-PER' but Softmax don't know. Compared to Softmax layer, CRF layer could use sentence-level tag information and model the transition behavior of each two different tags.

dataset

#sentence #PER #LOC #ORG
train 46364 17615 36517 20571
test 4365 1973 2877 1331

It looks like a portion of MSRA corpus.

train

python main.py --mode=train

test

python main.py --mode=test --demo_model=1521112368

Please set the parameter --demo_model to the model which you want to test. 1521112368 is the model trained by me.

An official evaluation tool: here (click 'Instructions')

My test performance:

P R F F (PER) F (LOC) F (ORG)
0.8945 0.8752 0.8847 0.8688 0.9118 0.8515

demo

python main.py --mode=demo --demo_model=1521112368

You can input one Chinese sentence and the model will return the recognition result:

demo_pic

references

[1] Bidirectional LSTM-CRF Models for Sequence Tagging

[2] Neural Architectures for Named Entity Recognition

[3] Character-Based LSTM-CRF with Radical-Level Features for Chinese Named Entity Recognition

[4] https://github.com/guillaumegenthial/sequence_tagging

About

A simple BiLSTM-CRF model for Chinese Named Entity Recognition 中文命名实体识别 (TensorFlow)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 66.8%
  • Perl 33.2%