Missing RDF triples detection and correction in knowledge graphs

Abstract

Knowledge graphs (KGs) have become a powerful asset in information science and technology. To foster enhancing search, information retrieval and question answering domains KGs offer effective structured information. KGs represent real-world entities and their relationships in Resource Description Framework (RDF) triples format. Despite the large amount of knowledge, there are still missing and incorrect knowledge in the KGs. We study the graph patterns of interlinked entities to discover missing and incorrect RDF triples in two KGs - DBpedia and YAGO. We apply graph-based approach to map similar object properties and apply similarity based approach to map similar datatype properties. Our propose methods can utilize those similar ontology properties and efficiently discover missing and incorrect RDF triples in DBpedia and YAGO.