A Concept-Relation Vector Model Based Method for Web Document Retrieval

Semantic information has been paid much attention in web IR. Although many researches have improved the retrieval performance by employing WordNet synset and concept, relations between concepts are often ignored by most of the semantic retrieval methods. We propose a relation enhanced concept vector model CRVM(concept-relation vector model) for document representation in this paper, and the documents to be retrieved are indexed by both concepts and relations. Domain ontology is employed to provide background knowledge for constructing concept based vector representation of documents. We prove the effectiveness of ontology concept and relation enhanced document representation for retrieving by web pages derived from WebKB data set and Open Directory Project.