Chinese Name Entity Recognition Using Highway-LSTM-CRF
Dongyang Zhao, Jiuming Huang, Yan Jia · 2018
We have implemented a highway-LSTM-CRF(Long Short-Term Memory, LSTM for short; Conditional Random Field, CRF for short) model for Chinese NER(Named entity recognition, NER for short), which encodes a sequence of input characters as well as all potential words that match a lexicon. Through the highway layer, our model can intelligently select words that are more relevant to the current character. In this way, an effect similar to the attention mechanism is achieved. Our model uses word and word sequence information without being affected by word segmentation errors. Experiments on various datasets demonstrate the effectiveness of leveraging lexicon knowledge and the efficiency of our model.