Research On Rlaxation Of Fuzzy Query On RDF

With more and more diversified contents on the WWW, people's desire to accurately, fast, and fully gain information is becoming stronger and stronger. As the extension of the current Web, the semantic Web researches the addition of semantics which can be read by machines to the WWW, thus the machines can gain comprehensive and useful information fast and accurately from the WWW and automatically process them. RDF (Resource Description Framework), as a common metadata model standard, makes the semantic retrieval in the semantic Web possible.There are many query languages in the semantic Web, which supports the RDF query, and now SPARQL has become the recommended standard of the RDF query. Still, in order to satisfy the users'intention of fuzzy query, there comes the query language with the support of fuzzy query like the f-SPARQL, which makes the RDF query system possess fuzzy queries directly. However, because the users don't know the structure and content of the data set of RDF in the semantic Web, initial fuzzy query condition committed by users is likely to be so rigorous that there are too few results or none result returned, which greatly reduces the recall ratio of the fuzzy query. The users hope that the query system can return approximate and ordered query results automatically, while the current systems of fuzzy query don't have this function. As to this, this thesis provides the relaxation of fuzzy query on RDF and a solution of ordering of the result set after query relaxation.The core content of this thesis is:the research on relaxation of atomic fuzzy query and compound fuzzy query based on RDF fuzzy query, and the acquirement of the result set. The beginning of the article is about the relaxation algorithm of atomic fuzzy query which fails to return enough results on RDF data set, whose core technology is relaxation operations of fuzzy terms and the determination of tolerance indicator Z, and about how to make the query relaxation iteration a controlled process by using non-authorized fuzzy sets of query field. Also, this thesis researches complex relaxation of compound fuzzy query of RDF. Here IDF weight information and relaxed query lattice structure of fuzzy query condition determines the execution order of many different relaxed query patterns. In order to reduce system execution time, the minimum failure subquery is used in the relaxation iterative process to acquire an efficient and controllable relaxation algorithm. Still, this thesis comes forward with a general scoring strategy based on similarity degree of fuzzy sets and weight of query condition. Ranking of results of relaxed queries which is based on TA algorithm can efficiently complete the ordering of results after relaxation. We design the prototype system of the RDF fuzzy query which contains a self-adapting relaxation function and the analysis of its experimental results. Experiments show that the relaxed results have good performances on recall and precision, but the response time of our system should be improved.