An approach of entity alignment based on semantic features

Entity alignment is to link “equivalent” entities that denote the same object in the real world, which can be used to help promoting information integration and retrieval. Driven by the knowledge graph initiative, large amount of owl:sameAs links need to be established. Most of the existing entity alignment methods are based on the matching progress of schema pattern, and there are limitations in dealing with the alignment process in Linked Open Data datasets, lots of missing entity links is still to be found. In this paper a schema-independent entity alignment method based on attributes semantic features is proposed. In the feature modeling approach presented here, attributes of entity in the Linked Open Data datasets is modeled according to the semantic features similarity integrated with label statistical characteristics. Finally, we use a supervised variable sets classification algorithm to realize classifier optimization. The experiments show that the method is more efficient and has better effect in accuracy, recall and F measurement.