A Logic-Agent based System for Semantic Integration

The heterogeneity of information sources can syntactic or semantic. The former can be due, for example to platform mismatches and different Database Management Systems, and, the latter to different database schemas and ontological mismatches [1]. Syntactic heterogeneity is catered for by middle-ware components known as wrappers, which provide a relational view of the data stored in the sources. The semantic integration of information involves providing a uniform interface to a number of semantically heterogeneous information sources. This is achieved via a global schema (a set of views) and a global ontology, in the spirit of the mediator architecture introduced by Wiederhold [2]. The users can then form queries using the global schema, independently of the schema (and ontology) of the sources. If the views that consist the global schema are materialized, the information integration task reduces to Datawarehousing. Data Warehouses provide users with very fast answers, since all the available information is stored locally, but the cost of updating them when the underlying sources change can be very high [3]. However, if the global schema is virtual, then the user query has to be translated into another query or possibly a set of disjunctive (alternative) queries (known as query-plans) that can be posed to the information sources “on the fly”. There are two main approaches to the virtual semantic integration task, namely the query-centric approach and the source-centric one. In the first, the views of the global schema are defined in terms of the information sources, reducing the query-translation to unfolding the definition of each view of the global schema occurring in the user-query [4]. The disadvantage of this approach is that if an information source becomes available, is altered or is no longer available, the definition of the global schema has to change. One possible way to overcome this is to “reverse” the definition of the global schema, which is the idea behind the source-centric approach. In this approach, each source provides the mediator with a description of itself, in terms of global schema views. These descriptions are then used to find the relevant sources, when a user-query is posed [5]. The architecture of the system in this case is much more flexible, but the query-translation task is much more complicated; the two most well-known algorithms for this task are the inverse-rules and the bucket algorithm [6]. Many information integration systems use AI ideas, mainly knowledge representation techniques, planning and learning. Most of the recent information systems in particular follow the agent paradigm, in order to have a more dynamic architecture.