Research And Implementation Of Retrieval System Based On Domain Ontology

Internet is a massive library of information resources,with the advent of a new era of big data network information,the amount of information is at an explosive rate of growth, and the organization of information is heterogeneous, multiple and distributed, how to find the information which meets the needs of the user's query expectation in the information resource library has become a major challenge for the current information retrieval system. This paper make in-depth studies on the information retrieval based on domain ontology, and propose a new semantic mapping model, then study about the construction of the domain ontology, document similarity matching and ontology query expansion are developed as following:1. The construction of Domain Ontology. According to the principle of "Seven Steps" of Stanford University, this paper puts forward a method of constructing domain ontology and gives an example of a specific domain ontology by ontology development tool Protege.2. Propose a new semantic mapping model. Based on the linear chain conditional random fields, an improved hidden dynamic conditional random field is used to establish the mapping model between ontology and vocabulary, and combined with the context words, to store the mapping relationship between ontology concepts and vocabulary,so as to achieve the purpose of word sense disambiguation.When the user query and the documents in the document library are transformed into the conceptual vector after the semantic mapping model disambiguation, then using vector space model to calculate the semantic similarity and the matching document set is returned according to the similarity from large to small. The contrast experiment shows that Hidden dynamic conditional random field is better than Hidden Markov model and Maximum entropy model.3. Research on ontology-based query expansion. Because the domain ontology can describe and define the relationships among concepts, and has strong semantic expression ability, when the user's query is mapped to the corresponding concept in the ontology, it can also be extended to improve the retrieval efficiency . The method is based on Ontology class axiom, using the custom inference rules of Jena reasoning mechanism to achieve ontology query expansion, including class relation reasoning,class / instance relation reasoning and attribute-based reasoning.Based on the above research, this paper develops an information retrieval system based on sports domain ontology, which supports the keyword-based retrieval,ontology-based extension retrieval and ontology-based semantic retrieval. The experimental results show that the ontology-based semantic retrieval is better than the keyword-based retrieval and ontology-based extension retrieval, therefore, it can meet the needs of people's higher retrieval,and provide a basis for the further study of semantic information retrieval.