Web, Semantics, OIL and FUEL:
Semantic Interoperability and learning on the Web

 Abstract for
Stanford DB Seminar,
October 20, 2000

Amit Sheth

Director, LSDIS Lab, Univ. of Georgia; Founder, Taalee, Inc. amit@cs.uga.edu




The concept of Semantic Web juxtaposes semantics and the Web.  Semantics (the meaning and use of data) brings information closer to human thinking and decision-making.  As we did in the pre-Web era, we are now again going to investigate different forms of semantics (patterns, domain modeling, logical and semantic relationships among homogeneous or heterogeneous and  real or virtual/inferred entities, etc) from the perspectives of different scientific disciplines (information and knowledge-based systems, AI, linguistics, cognitive science, etc.). The scope of the Web now forces us to simultaneously deal with the complexity of modeling and reasoning, with the huge scale and heterogeneity of all imaginable kinds.


Despite these challenges of heterogeneity and scale, and despite the relatively modest commercial success of federated and mediator systems, we have started to see large-scale, industrial strength, albeit early, creation of the Semantic Web for commercial applications. Take three examples—Oingo [2], GuruNet and Taalee [3].  The latter, for the first time, provides semantics-based, pan-Web services for categorization, cataloging, search, directory, personalization, targeting, and more.  This commercial activity is complemented by standardization and research activities related to DAML and OIL.


In this talk, we will focus on more complex problems and the desire to support a what-if based learning approach in the context of the Semantic Web.  In particular, we discuss a mechanism for semantic information correlation, which provides a framework for semantic interoperability.

We have called this information correlation [1], MREF (metadata reference link) in our InfoQuilt project and Information Landscape in ADEPT [4], our early version of a Digital Earth prototype system that is a part of Digital Library II program.  To support “deep semantics” and a paradigm for learning, we go beyond standard relationships supported in logic-based frameworks, and define an “affects” relationship.  An issue of future investigation is how can we FUEL semantics-based organization of the Web and the corresponding reasoning mechanisms, building upon DAML-OIL types of specifications and

logic-based reasoners.



Further reading


[1] V. Kashyap and A. Sheth, Information Brokering, Kluwer Academic Publishers, 2000

[2] Oingo, http://www.oingo.com

[3] Taalee, http://www.taalee.com

[4] ADEPT project and prototype at UGA (reports, presentations and student theses), http://lsdis.cs.uga.edu/~adept