Research On Semantic Retrieval Model Based On Ontology

For the sake of getting perfect and valuable information along with the rapid development of network and the surge of information on it, people pay more and more attention on Retrieval Technology. Information Retrieval System, as an indispensable part of network, always plays an important role during the process of meeting users' requirement.Traditional Search Engines which is concerned about matching key words, doesn't involve the level of semantic. There are some problems like incomplete expression, different expression, and word's isolation and so on. It gives users much trouble for the sake of its numerous search results. Recall and precision of those engines are not able to fulfill users' demand. In response to the features above, great importance has been attached to semantic retrieval technology and it has become hotspots of research. Raising and developing of Semantic Web provides a basis for Semantic Retrieval, and at the same time, semantic retrieval includes many concepts and instances. Because of good concept hierarchy, Ontology is of important value to the information retrieval.For these reasons, we give elaboration of speculative knowledge and theory of Ontology, and then introduce the concept and architecture of Semantic. We propose the model of semantic retrieval based on Ontology. It's distinct from traditional models which are based on keyword matching, because it is based on concept matching. It returns documents that are match searched concept. And it's different from some existing search methods making use of deduction through Ontology, the new model reduces dependency of Ontology's integrity, it's suitable for the internet rather than a specific domain. There for, it is much more practical and wide range of application.In the course of building Ontology, Skeletal Methodology Principle was used. After word segmentation and statistics of frequency, we got the Ontology of Computer Domain which contains concepts and instances and relevance of them. Weight calculation of relevant words between concepts is enhanced while computing weight of concepts. Representative value is represented by word similarity, and we set document's similarity by comparing Document-Concept Weight which is different from algorithm of Vector Space.We experiment by structured Ontology, and we get improvement both in the scopes of precision and recall comparing with traditional search engines of different similarity. Thus the model of semantic retrieval based on Ontology is better than traditional models.Semantic retrieval based on Ontology is developing according to Semantic Network, Ontology, and natural language processing. There is huge research space of building of Ontology, combination between Ontology and semantic retrieval, calculation of word similarity and evaluation of searching result, those are the next major job.