VEGA Project

Knowledge Validation and Exploration by Global Analysis

The VEGA research tries to provide a structured framework for the VALIDATION and EXPLORATION of knowledge bases (KBs) by global analysis. This knowledge EVOLUTION framework will be offered to developers and users of KBs.

VEGA Staff:

Dr. Harold Boley (boley@informatik.uni-kl.de), project leader
Andreas Abecker (aabecker@dfki.uni-kl.de)
Dr. Knut Hinkelmann (hinkelma@dfki.uni-kl.de)
Otto Kuehn (kuehn@dfki.uni-kl.de)
Dr. Franz Schmalhofer (schmalho@dfki.uni-kl.de)
Holger Wache (wache@informatik.uni-bremen.de)

Intended Functionality:

While much knowledge is acquired and represented in KBs, often its interconnections remain unexplored (lack of discovery support) and its content is not validated (lack of quality control). Expert systems accessing such KBs suffer from the resulting brittleness. Knowledge engineers shall thus interact with the VEGA system for evolving (i.e., exploring and validating) KBs represented in a highly portable (declarative) language.

AI Tasks:

Knowledge exploration: Detect interesting connections between items in a KB.
Knowledge validation: Increase a KB's trustworthiness and performance.

Solution Approach:

Provide a language for both problem solving and KB evolution, representing knowledge at a high level of abstraction, and translators from existing KBs to that language.
Supply techniques for KB exploration and validation, combining global-analysis and knowledge-transformation methods with scalable, symbolic machine-learning mechanisms.
Develop a methodology for exploring hypotheses formulated in the KB language and feeding back interesting, validated patterns into the KB.

Application Area:

The VEGA approach shall be demonstrated with KBs supporting engineers. In particular, we are building a KB to be used for recycling-oriented product and production planning in the domain of plastics and composite materials.

The VEGA project is divided into three parts:

DRL: Declarative Knowledge Representation Language (here RELFUN)
KES: Knowledge Evolution System (here currently available in German only)
APPLKB: Application Knowledge Base (here RPPP)

VEGA's joinder of sharing and evolution is found in a German review (here); more details and publications in its German-English final report (here).

contact: webmaster@dfki.uni-sb.de / last change: 26 Mar 98