A Method of Ranking Structured Queries for Keyword Search on Semantic Web Data

To find information from Semantic Web, users need to use structured query languages like SPARQL. Using those languages, a user can express her information need exactly, but the usage is not familiar to ordinary users. To help the users find information easily, the methods of converting keyword queries issued by users into structured queries for Semantic Web are being researched. Keyword queries are easy to use, but it has ambiguity problem, in other words, a keyword query can be converted into a great number of structured queries. Each structured queries converted from a keyword query have different importance and relevance to the keyword query, thus, the ranking method for those structured queries is needed. In this paper, we propose a ranking method for structured queries converted from the same keyword query, which uses the affinities between schema-level triples in Semantic Web data. The proposed method extracts the affinities from the query log, and its ranking function for structured queries is based on the affinities. Because the proposed method reflects the preference of users, it can rank the structured queries more effectively than ranking methods using only the information of target domain ontology.