demianzhang/ChineseNER

A neural network model for Chinese named entity recognition

Recurrent neural networks for Chinese named entity recognition in TensorFlow

This repository contains a neural network model for chainese named entity recognition.

Contributer

• Jingyuan Zhang
• Mingjie Chen

Requirements

• Tensorflow
• jieba

Model

The model is a birectional LSTM neural network with a CRF layer. Sequence of chinese characters are projected into sequence of dense vectors, and concated with extra features as the inputs of recurrent layer, here we employ one hot vectors representing word boundary features for illustration. The recurrent layer is a bidirectional LSTM layer, outputs of forward and backword vectors are concated and projected to score of each tag. A CRF layer is used to overcome label-bias problem.

Basic Usage

Default parameters:

• batch size: 20
• gradient clip: 5
• optimizer: Adam
• dropout rate: 0.5
• basic learning rate: 0.001

We train word vectors with gensim version of word2vec on Chinese WiKi corpus and finetune the embedding layer

Train the model with default parameters:

$ python3 train.py

the best F1 score is 0.8948 when tested, a f1 score higher than 0.89 is reasonable