Research And Implementation On Query Expansion Model Of Information Retrieval Based-on Conceptual Graph

The current semantic-based information retrieval systems identify some semantic information in a degree. However, because these systems are so lack of a degree of language analysis and understanding and their query and resources expression is a so discrete analysis that Words lost contact between the intrinsic. Based on those facts the retrieval precision are declined.In order to improve the precision rate and recall rate in information retrieval systems,this paper adopts conceptual graph theory in natural language processing and "HowNet" semantic dictionary and propose and implement a semantic search query expansion model of information retrieval systems.This paper mainly include follow aspects:(1)Query expansion algorithm. At present, the query expansion of information retrieval are based on statistics. Combining statistics method and "HowNet", a method based on "HowNet" is proposed,which achieves the similarity of new documents and query words with Re-weighting word items by N-level Vector Space Model, to improve the effectiveness of retrieval systems.(2)Generation of concept graphs. Generation of concept graphs is a process to format semantics in linguistics, which is a first step for computer to understanding and also a base for all techniques of semantic. So-called formation in Linguistics is to give a strict and normative representation of the problems, which need to study , in some form of mathematics. Conceptual graph is adopted as a knowledge representation tool in this paper. We depend on syntactic tagging system of IR-Lab in Harbin Institute of Technology, and generate conceptual graphs by transforming from grammar relation to semantic relation according to characteristics of Chinese grammar,and realize mathching of conceptual graphs.(3)Designed and implemented a concept graph based query expansion model of information retrieval systems. This system mainly includes query expansion module and concept graph matching module. The former ensures that the same concept expressed in different words can be retrieved and the latter is the key factor for retrieving the same or very similar words or phrases. This paper adopts query expansion techniques to retrieve related documents and to improve recall in information retrieval. This paper also improve precision of information retrieval by projection matching, the biggest connection matching and similarity computing of conceptual graphs to retrieve matched documents.Finally, The results of the evaluation system are given. Query expansion techniques and the method used in conceptual graphs generation and conceptual graphs matching is tested and anlysised. Comparison of the result with statistical based model is carried out in a experiment and experimental result indicates that the semantic search query expansion model has improve recall and presion of information retrieval system.