Attention-based Recurrent Neural Model for Named Entity Recognition in Chinese Social Media
Binhui Wang, Yanyu Chai, Shusong Xing · 2019
NER (Named Entity Recognition) is one of the major tasks in NLP (natural language processing). Although there are already some methods to solve NER problem, it is rare in Chinese social media, and less of them can make good use of external information. The POS (part-of-speech) of a word can convey a lot of information in NER, and the attention mechanism can help us successfully focus on these contents. For the above reasons, we propose an attention-based bidirectional LSTM (Long Short-Term Memory) model to address NER problem in Chinese social media. The attention mechanism utilizes the POS of every character of a sentence in the learning process. Experiments on the dataset of Chinese social media demonstrated the performance of the proposed neural model.