Research on power grid contract entity and relation extraction based on domain embedding

Power grid contracts have complex structures and domain specific entity relationships, posing challenges to natural language processing models. This paper proposes a BERT BiLSTM CRF entity and relationship extraction model based on domain embedd ing to better capture relevant entities and relationship information in power grid contract texts. Firstly, we construct a domain embedding matrix based on contract text information, generate word vectors representing contextual semantic information throug h the BERT preprocessing model, and add domain embedding vectors to the output of BERT. The generated vectors are then feature extracted through BiLSTM, and the entity relationship sequence is decoded through CRF. The purpose of this paper is to improve th e accuracy of the model's extraction of entity relationships in power grid contracts through domain embedding, laying the foundation for subsequent contract text standardization, risk prediction, and intelligent review. Experiments have proven that the mod el can effectively identify entities and relationships in power grid contracts.