Relation Identification between Name Entities Based on Community Structure Mining

The co-occurrence of the name entities in sentences and documents usually implies some important relationships among them. This paper addresses relation extraction problem and proposes an unsupervised method of automatic discovering relations among entities based on community structure mining. After distilling all named entities pairs, a network is constructed to represent the semantic relationship among name entity pairs and a mining strategy is employed to adaptively analyze the network and detect the different relation communities. Finally, each community is labeled by an indicative word. Experiments show that our method performs well on both high-frequent and less frequent entity pairs, at the same time appropriate labels could be automatically provided for the relations.