Abstract
he electric power business is characterized by numerous types, wide audiences, and a high level of expertise, so we need to develop intelligent customer service for the electric power field based on the domain-specific knowledge graph. In this paper, a knowledge graph was constructed for the electric power business according to the research on the intelligent dialogue system for the electric power field, based on the electric power expertise and the daily service Q&A scenarios of Anhui Electric Power Service Hall. We proposed a construction method that consists of the knowledge modeling method for building a knowledge graph for the electric power business, knowledge extraction method based on machine learning, triplet knowledge representation method, and knowledge storage method based on a graph database. Multiple entity and relation extraction methods were compared through experiments. The BERT-BI-LSTM-CRF and BERT-BI-LSTM-ATT methods that perform the best in experiments were selected for knowledge extraction, and 4,109 entities and 1,587 triplets were extracted.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call