The Character Relationship Mining Based on Knowledge Graph and Deep Learning
Ying He, Hongyan Yun, Lin Li · 2019
Character relationship information is important for precision poverty alleviation. The main challenge is that the character relation data is heterogeneous and relation words are implicit in text short sentences. Therefore, we propose a method of character relationship mining based on the combination of knowledge graph and deep learning. The DNN and BiGRU neural network joint method are proposed to recognize person named entities and extract character relations. The character relation triples data are stored in the Neo4j graph database, and the deeper information is mined by using the multi-depth relation query method. The experimental results on the poverty alleviation data show that the best accuracy of relation extraction can reach 84.9%, and the query efficiency of character relationship is improved. Users could quickly view the implied character relationships through the graph.