Back to Main Semantic Integration Through Invariants (Presentation)

    Joseph B. Kopena+* and Michael Gruninger-
   tjkopena at cs dot drexel dot edu, gruning at cme dot nist dot gov

+    Geometric and Intelligent Computing Laboratory
     Department of Computer Science
     College of Engineering, Drexel University

-    Manufacturing Systems Integration Division
      Manufacturing Engineering Laboratory
      National Institute of Standards and Technology

(* at the National Institute of Standards and Technology during this work and presentation preparation)

   Software and knowledge integration presents a daunting challenge for a world increasingly driven by and reliant on information technology. A body of computer applications and code of ever-growing size and complexity makes connecting software to solve new problems progressively more complicated, resource-consuming, and error-prone even as it makes doing so more rewarding. Similarly, the extensive, diverse, and expanding corpus of knowledge, information, and data available on the World Wide Web promises to raise automated agents and services to a new level of utility and capability if it could be rendered software accessible. However, reasonable approaches to attributing machine-interpretable semantics to Web content will necessitate integration of those representations.

   One approach to resolving these issues is interoperation through an interlingua---exchange between integration targets is accomplished through translations to and from a shared language. By using a formal ontology as the interlingua, the targets can be characterized using invariants rigorously defined by properties of possible models of the interlingua. These characterizations leverage the structure of the domain as defined by the interlingua to enable such tasks as providing unambiguous definitions of target concepts, performing automated translation at a semantic level, compiling point-to-point mappings, coverage and interoperability analysis, and developing formal ontologies for informally specified legacy systems.

   This talk presents work on generating these characterizations, developing retrofit ontologies, comparing and analyzing integration targets, compiling direct mappings, and translation through an interlingua framework based on these characterizations. Attention is paid throughout this work to supporting applications and information sources with and without an existing explicit formal ontology. Examples and demonstrations are drawn from the domain of business and manufacturing processes. The NIST Process Specification Language (PSL) is used as the interlinqua. Representative targets include: the process modeling component of DAML-S, a web services description language; IDEF3, a process modeling language widely used in industry and business; and ILOG, a commercial process planner and scheduler.

   The presentation slides have been posted for download [ pdf, 1.4M ].

   Presentation given at National Institute of Standards and Technology (NIST) Summer Undergraduate Research Fellowship (SURF) Closing Presentations Plenary Session (Manufacturing Engineering Laboratory Representative); August 11 2003; Gaithersburg, MD. - Joe Kopena updated://wednesday/2-11-04?8.11.5