Research on Power Information Knowledge Extraction Model Based on BERT
Zha Yiyi, Chong Wang, Mingming Zhang, Kai Liu · 2021 IEEE Sustainable Power and Energy Conference (iSPEC) · 2021
Knowledge extraction is the basic step of constructing power information knowledge graph. Aiming at the problems of polysemy, poor context awareness and limited sentence types in the previous text processing methods in the power field, we propose a knowledge extraction method based on BERT from the characteristics of power information text. In the first part, BERT-BiGRU-CRF model is used to complete the named entity recognition of power information text. In the second part, BERT-BiGRU-attention model is used to extract the relation between entities. The model can improve the effect of named entity recognition and relation extraction in complex context, and can process diversified power information texts.