Answering queries in context

Abstract

The emergence of the Internet as the de facto Global Information Infrastructure enables the construction of decision support systems that leverage on the panoply of on-line information sources. This highly dynamic environment presents a critical need for a flexible and scalable strategy for integrating the disparate information sources while respecting their autonomy.

The Context Interchange strategy addresses above concerns with an emphasis on resolving problems arising from semantic heterogeneity, i.e. inconsistencies arising from differences in the representation and interpretation of data. This is accomplished using three elements: a shared vocabulary for the underlying application domain (in the form of a domain model), a formal object-relational data model (coin), and an object-deductive language (coinl). Semantic interoperation is accomplished by making use of declarative definitions corresponding to source and receiver contexts; i.e. constraints, choices, and preferences for representing and interpreting data. The identification and resolution of potential semantic conflicts involving multiple sources are performed automatically by the context mediator. Users and application developers can express queries in their own terms and rely on the context mediator to rewrite the query in a disambiguated form.

This work is supported in part by ARPA and USAF/Rome Laboratory under contract F30602-93-C-0160, the International Financial Services Research Center (IFSRC), the PROductivity in Information Technology (PROFIT) project at MIT, and ARF RP970628 at the National University of Singapore.