| Summary: | In the last years, users have become used to keyword-based search interfaces due to their ease of use. By matching input keywords against huge amounts of textual information and labeled multimedia files, current search engines satisfy most of users'' information needs. However, the principal problem of this kind of search is the semantic gap between the input and the real user need, as keywords are a simplification of the query intended by the user. Moreover, different users could use the same set of keywords to search different information; even the same user could do it at different times. The search system, before accessing any data, should discover first the intended semantics behind the user keywords, in order to return only data fulfilling such semantics. The use of formal query languages is not an option for non-expert users, so a semantic keyword-based search based on semantic interpretation of keyword queries could be the solution, i.e., a search that starts discovering the semantics intended for the input user keywords, and then only data relevant to that semantics are returned as answer. In this paper we present a system that performs semantic keyword interpretation on different data repositories. Our system (1) discovers the meaning of the input keywords by consulting a generic pool of ontologies and applying different disambiguation techniques; (2) once the meaning of each keyword has been established, the system combines them in a formal query that captures the semantics intended by the user, considering different formal query languages and possibilities that could arise, but avoiding inconsistent and semantically equivalent queries; and, finally, (3) after the user has validated the generated query that best fits her/his intended meaning, the system routes the query to the appropriate data repositories that will retrieve data according to the semantics of such a query. Experimental results show the semantic interpretation capabilities and the feasibility of our approach.
|
|---|
