Chinese Conference Event Named Entity Recognition Based on BERT-BiLSTM-CRF
Rui Xiong · 2020
Conference events are an important place for people to express their views. Chinese conference event named entity recognition is a key technology for public opinion tracking of people. In this paper, a method based on BERT-BiLSTM-CRF is proposed to recognize Chinese conference event named entities. First, a character vector is generated by the BERT model based on context information. And then the trained character vector is input into the BiLSTM-CRF model for further training processing. The experimental results show that BERT-BiLSTM-CRF model performs better than classical BiLSTM-CRF model and word2vec-BiLSTM-CRF model. BERT-BiLSTM-CRF model can achieve precision of 91.39%, recall of 92.31% and F1-score of 91.85% in Chinese conference events named entity recognition.