/project001

hum...

Primary LanguagePythonApache License 2.0Apache-2.0

NER: Chinese Named Entity Recognition in Keras

Model1:Character_embedding-Based BiLSTM-CRF.
Model2:On the basis of model1. Extract n_gram feature from word embedding as auxiliary feature, using Conv1D.

Ps:Run preprocess.py and utils.py firstly, to get train、dev and test data.

References:
1.End-to-end Sequence Labeling via Bi-directional LSTM-CNNs-CRF. Xuezhe Ma, Eduard Hovy
2.Bidirectional LSTM-CRF Models for Sequence Tagging. Zhiheng Huang, Wei Xu, Kai Yu.
3.Neural Architectures for Named Entity Recognition. Guillaume Lample et al.