I will describe the Information Manifold system, developed at AT&T
Bell Labs, that provides a uniform query interface to large
collections of structured information sources. The Information
Manifold combines novel techniques from the two fields of Artificial
Intelligence and Database Systems. Specifically, the Information
Manifold provides a rich language for describing both the contents and
the capabilities of information sources, and algorithms for
determining precisely which information sources are relevant to a
given query.
The representation language used in the system is based on CARIN, a
novel combination of Horn rules and Description logics, two formalisms
that are necessary for this application. I will describe the core
results about this languages, which include a sound and complete
reasoning procedure for non-recursive CARIN-ALCNR knowledge bases, and
an analysis about when exactly reasoning is decidable in the presence
of recursion.