EXPECT: Intelligent Support for System Construction and Documentation
Quarterly rept. 12 Sep-11 Dec 1991.
UNIVERSITY OF SOUTHERN CALIFORNIA MARINA DEL REY INFORMATION SCIENCES INST
Pagination or Media Count:
During the period, our effort concentrated on further integrating the EES framework, on which EXPECT is based, with the LOOM knowledge representation system, on getting acquainted with the TRANSCOM domain, and on improving the text plannar that will be used in this project. We have now completed the specification and re-implementation of how goals are to be matched to plans. As we now make use of the LOOM classifier capabilities, our goal matching algorithm is more powerful and faster than the original algorithm in EES, and the organization of plans in the knowledge base more compact. More specifically, plans capabilities and goals are all represented in LOOM, thus yielding to an expressive and uniform representation. The matching of goals to plans is done with the LOOM classifier, which allows for efficient semantic matches, resulting in a more robust and efficient system. The matching of goals to plans makes use of the knowledge in the domain model. Such semantically oriented approach to this matching problem together with the resulting plan organization is a new approach in planning. It is described more fully in the first of the attached technical reports Organizing Plan Libraries in Subsumption Hierarchies Specificity Based Plan Selection.
- Computer Programming and Software