Conceptual Representing of Documents and Query Expansion Based on Ontology

摘要

In vector space model, a document is represented by words. As the new words appear dramatically in the Internet era. this kind of method draws back the IR systems performance. This paper puts forward a new approach to present the concepts, query expressions, and documents based on the ontology. The approach has two levels, the Word-Concept level and the Concept-Document level. In the first level, the transition probability matrix is constructed by using the appearing times of word-word pairs in documents. The biggest eigenvector of matrix is computed, and it reflects the importance of words to the concept. In the second level, the distance matrix is constructed by using the distance between words in a given ontology, and the average variance value of elements is computed. It reflects the relevance of documents to concepts. In the last section, the query expansion is discussed by using the personal information profile of the user. It is proofed to be more effective than previous one.