A Clustering-Based Infrastructure for Reuse of Ontologies
  • Agency: Office of Naval Research. SBIR Phase I & II
  • Duration: September 2003 – January 2008.

   Capabilities Being Demonstrated:
  • Comprehension of Large Complex Semi-structured Systems
  • Reuse Opportunities across Knowledge-Based Systems
  • Mapping and Alignment Support for Multiply-Authored Knowledge-based Systems

Currently, aligning, mapping and merging knowledge-based systems are complicated undertakings that require significant manual effort from their designers. Three primary barriers to effective reuse are:

Where is it? Due to the number and size of existing knowledge-based systems that are candidates for reuse, finding a concept that can be referenced or adapted in a context of interest is a daunting task. Different KBSs support different representational features, use different terminology, and structure their models differently, making simple search mechanisms ineffective.

How is it used? To effectively reuse a concept, a knowledge base designer must understand the contexts in which the concept is defined and used in the original KBS. Importantly, concept definitions do not exist in isolation. The manner in which the concept is used—typically, in constraints and/or inference rules—is at least as important as the manner in which it is defined. Pragati’s analysis of numerous production knowledge bases has shown that concepts very often collaborate in interesting and non-obvious ways. These collaboration patterns become apparent only upon examination of the concepts’ usage. For large knowledge bases developed by others, finding collaboration patterns to support effective reuse by hand is extremely difficult. However, we believe reuse depends on this pillar of discovering collaboration patterns.

How to adapt it? Real-world reuse situations are often very complex. Many knowledge-based systems define concept representations that are broadly similar but whose details differ, possibly at the syntactic level, and almost certainly at the semantic level. The differences may be due to authoring bias, varied shades of meaning in the real-world concept upon which the KBSs’ formal representations are based, or different orientations of the defining systems. For a KBS builder who desires to pull aspects of several different existing concept representations into a new concept definition, reuse becomes an exercise in copy-and-paste frustration. Very often, a knowledge-based system containing domain knowledge that should be highly reusable in a target context is effectively inaccessible.

Our ONR-SBIR project addresses development of a knowledge engineering environment, Expozé, that will enable ontology builders to find and reuse relevant portions of existing ontologies in interesting and intuitive ways. It is based upon Pragati’s core technology, MVP-CA (Multi_ViewPoint Clustering Analysis).

« Back | Top
©2007 Pragati Synergetic Research, Inc. Research Focus | Products | Projects | Publications | Market Sectors | About Us