Università Degli Studi Di Modena E Reggio Emilia

Query Processing in mediator-based systems primarily deals with query answerability. Referring to the integrated information sources, content and query capabilities description are then essential to successfully answer a query. Moving to Multi-Agent Systems, a query must be compiled into an information request to be fulfilled by some information provider. Both the SEWASIE Brokering Agent (BA) and the DAML-S Broker (DSB) are middleagents involved in query management and share some open issues. In particular, they must embody matching functionalities to discover an information provider capable to answer a given information request. Using DAML-S as a capability description language let the DSB behave as a semantic Web service, thus achieving a Semantic Matching between an information requester and relevant information providers. Within this framework, DQL (DAML Query Language) has been advised as the query language for the DSB. Query Processing requires a substantial Query Reformulation process to enable Semantic Matching. Such a Semantic Reformulation is split into Query Abstraction and Query Rewriting. W.r.t. a common reference ontology both Query Abstraction and Query Rewriting have been formalized and implemented. The motivating rationale is to abstract instance or variable terms of the initial query into concepts representing service capabilities. Once a matching provider for such an abstracted query has been found, concept rewriting must be performed in order to supply the provider with the expected parameters. Intra-ontology relationships are exploited to accomplish such a transformation. Moreover, semantic similarity (WordNet aided) heuristics enhances the whole multi-phase procedure. Casting the DAML-S paradigm to the SEWASIE Brokering Agent might allow to keep the same strategy for Query Processing even for such a broker. Nevertheless, the subtask of Query Reformulation could be equally applied to any situation in which a query needs to be abstracted and rewritten in order to be successfully answered. A Description Logics based ontology language (i.e. OWL DL) supports the whole process. In spite of this, a proved exportability from/to ODLI3 permits to use it too for the reference ontology.