Attention in Character-Based BiLSTM-CRF for Chinese Named Entity Recognition

Abstract

Named Entity Recognition (NER) is a crucial step in natural language processing (NLP). Recently, some researches work on enhancing the word representations by character-level extensions in English and have achieve excellent performance. The same method applies to Chinese as well. In this paper, we introduce this idea into character-based Chinese NER. The method uses the information from character-level and glyph-level representations at the same time to achieve high accuracy. In addition, we propose an attentional model which is able to decide how much information to use from a character-level or glyph-level component dynamically. We validate the proposed method on the third SIGHAN Bakeoff MSRA datasets. Experimental results show that the proposed approaches are effective in improving the performance of Chinese NER and achieve 91.11% F1 without any additional handmade features.