Sharing co-reference knowledge for data integration

Data integration requires solving the basic problem of determining whether two identifiers from different terminologies co-refer, that is refer to the same object. The problem arises in many different contexts and provokes intense philosophical debate, yet it has never been addressed in a systematic way from an information system point of view. This paper presents a formal model of co-reference, as a relation which is part of the reality created by language structures. A complete mathematical characterization of co-reference is derived. A service handling co-reference knowledge is then introduced, and the basic functionality of such a service is formally defined in terms of a certain mathematical structure. Finally, the implementation of the service is studied from the computational point of view, and effcient algorithms for carrying out the basic operations of the service are provided.