Named Entity Recognition in Emergency Domain based on BERT-BILSTM-CRF
Yu Tian · 2022 IEEE 2nd International Conference on Electronic Technology, Communication and Information (ICETCI) · 2022
Chinese named entity recognition is a key problem in natural language processing. The traditional language processing model can not effectively represent the context semantic information in the text, and can not deal with the different contexts of polysemy, which affects the effect of entity recognition. A BERT-BILSTM-CRF entity recognition model for the emergency domain is proposed. BERT pre-training language model generates word vectors which represent the semantic information of context, and the generated vectors are characterized by two-way long-term and short-term memory network. Self-attention mechanism can effectively obtain the long-distance dependence in text sentences. And finally, decoding is carried out through the CRF to generate an entity label sequence. The experimental results show that the model have achieved excellent results in the emergency domain corpus.