Hierarchical Lexicon Embedding Architecture for Chinese Named Entity Recognition

Named entity recognition (NER) is one of the most fundamental tasks in a variety of natural language applications. Due to the lack of delimiters in the Chinese language, Chinese NER task has been suffering from the shortage of word boundary information. Recently, incorporating word information has been proven an effective mechanism to alleviate this problem. However, how to integrate word information into the character-based model more effectively and efficiently is still a challenge. In this work, we propose a hierarchical lexicon embedding architecture for Chinese NER task. The words matched by the input sentence are divided into two categories, i.e., main words and auxiliary words, to help the model better capture useful information. In addition, the modification mainly lies in the embedding layer, as such it can be easily incorporated with different sequence modeling architectures. Experimental studies on four Chinese NER datasets have shown our method’s promising potential.