Study On The RDF Fuzzy Query Method Based On Linguistic Variable

The Semantic Web is an extension to the existing World Wide Web, which attaches semantic meaning to Web resources and allows the computer to understand web information. With the development of semantic Web, the volume of RDF data used to describe Web data semantics grows rapidly. How to effectively acquire useful RDF information becomes the problem demanding prompt solution, so study on RDF query method has increasingly drawn people's attention.Traditional query methods cannot support RDF triple data retrieval. Researchers from numerous institutions or organizations have defined multiple query languages based on the characteristics of graph-structural data, these RDF query languages are important approaches of querying RDF data and acquiring the semantic information they expressed. There are dozens of common query languages, among which SPARQL language published by W3C is the most widely used language. At present, the RDF query method based on SPARQL language can support RDF exact query quite well. However, fuzzy query is sometimes used to produce approximate result when acquiring information in an actual application. Some work has already been done in RDF fuzzy query, but the research and application are mainly conducted under the framework of the database, which requires user to express fuzzy query request in a quantitative way. Fuzzy query methods for natural language are seldom involved in the study of current RDF fuzzy query. Based on fuzzy set theory, this study takes people's operating habits of daily query into consideration and takes large amounts of RDF data in the semantic Web as the object. Starting from the perspective of meeting ordinary user's fuzzy query needs and preferences and taking extended and standard SPARQL language as the prerequisite, this paper studies the characteristics of different types of RDF data based on linguistic variable. Fuzzy query methods of RDF in numeric type and RDF in text type are proposed. Characteristics of preference are analyzed in the paper. According to the unipolar and bipolar characteristics of preference information contained in multi-criteria fuzzy, RDF fuzzy query methods based on unipolar and bipolar preference are proposed; visual RDF query system that supports fuzzy criteria are constructed on the basis of the proposed methods. The main contents of this paper include:(1)To solve users' understanding problems caused by expressing fuzzy degree of query criteria with the membership degree, fuzzy query method of RDF in numeric type is proposed. Firstly, drawbacks of expressing fuzz qualitative issues using the quantitative way of membership degree are analyzed. The paper begins with fuzzy expression of linguistic variable and extends the classic SPARQL to support expression of the description of fuzzy criteria using the linguistic method; secondly, the fuzzy characteristics of absolute and relative quantifiers are analyzed in the light of RDF in numeric type. Fuzzy query methods for absolute and relative quantifiers are proposed, whiletransformation rules of fuzzy criteria are given based on fuzzy set theory; finally, the feasibility of the method are verified by constructing an experimental system, running typical examples and analyzing the results.(2)The critical problem of RDF fuzzy query in text type is the fuzzy match of keyword. This paper applies relative theories from TongYiCi CiLin (A Chinese Thesaurus) and CNKI comprehensively and proposes RDF fuzzy query method of extended keyword. Calculation formula of word distance is defined. Word distance is taken as the calculation parameter of word's semantic similarity; Formula is given to calculate word's semantic similarity through combination of the word sememes. The formula is used to calculate word's semantic similarity and words whose semantic similarity are greater than 0.5 are taken as the extended keywords, which are added to the keyword set. Hence, the fuzzy keywords for query are converted to exact keywords. Fuzzy query of RDF in text type is realized with the existing SPARQL query engine. Finally, experimental system is used to verify the feasibility of the method.(3)By analyzing the preference characteristics of preference information contained in daily expression, the preference of fuzzy query is classified into unipolar preference and bipolar preference and corresponding query methods are provided. Preference criteria in query are described in the linguistic method; extended SPARQL language supports preference expression. Weighted calculation of preference attributes is used to calculate the degree of preference for unipolar preference query. Preference query is implemented by calling the optimized results of sorting mechanism with the degree of preference as keyword. For bipolar preference query, RDF fuzzy query method of linguistic variable is used to get approximate result after defuzzification. The degree of results' compliance with bipolar preference criteria is evaluated using preference satisfaction function. A priority function is defined and results are optimized by sorting. Experimental system and corresponding experimental results verifies the effectiveness of the method.(4)With the increasingly widespread study of RDF, some query verification system appears. However, there is much less mature system for RDF fuzzy query. To implement user-centered operation, a RDF fuzzy query system is constructed. The system hides the complicated processing and fuzzy parameters hard for users to understand and support exact and fuzzy query quite well.Visual operation interface can help user implement selection query, group query and preference query more easily. "Select-entry" operation and SPARQL language are two query methods provided in order to satisfy different query habits of ordinary and professional users; manually adjusting the value of fuzzy match is set to help users adjust results according to their needs, meeting the demand for fuzzy query of different precisions.