Abstract: The most popular methods about information retrieval are based on keywords matching so far. Despite the search models which can indicate the meaning of documents and queries have emerged, they simply match the similarity of vocabulary semantically instead of lexically. However, the huge difference among existing ontologies and the various definitions of semantics with uneven qualities lead to the difficulties in semantic searching by user intention. Therefore, an efficient mechanism working on knowledge discovery and commonsense reasoning is extremely needed. This paper proposes a novel retrieval method combining ontology with graph theory. To infer the user intention, it presents the knowledge of documents using computer domain ontology and expands the queries in the shortest path of semantic network. In the search process, it ranks the documents based on both factors of lexical and semantic similarities. This model is proven efficient in improving the accuracy of retrieval through a testing of standard dataset.

Key words: semantic search, domain ontology, semantic network, query expansion