A graph-based approach for extracting terminological properties from information sources with heterogeneous formats
In this paper we propose a user profile based approach for supporting Web-oriented search. Differently from traditional syntax-based approaches, our method is semantic driven. Indeed, for each user, a profile is constructed and handled for representing her/his interests. Whenever a user submits a query, a classical Web search engine is activated for determining a (probably large) set of answer data sources. After this, the user profile is exploited for determining the subset of those data sources which best conforms to user interests. In order to semantically compare the content of a data source with a user profile, a formal inference framework, provided by a specialized probabilistic variant of Description Logics, is exploited.