Assessing Conceptual Similarity to Support Concept Mapping (pdf )

David B. Leake, Ana Maguitman, and Alberto Cañas. Proceedings of the Fifteenth International Florida Artificial Intelligence Research Society Conference. AAAI Press, Menlo Park, 2001, pp. 186-172. 5 pages.

Abstract

Concept maps capture knowledge about the concepts and concept relationships in a domain, using a two-dimensional visually-based representation. Computer tools for concept mapping empower experts to directly construct, navigate, share, and criticize rich knowledge models. This paper describes ongoing research on augmenting concept mapping tools with systems to support the user by proactively suggesting relevant concepts and associated resources (e.g., images, video, and text pages) during concept map creation. Providing such support requires efficient and effective algorithms for judging concept similarity and the relevance of prior concepts to new concept maps. We discuss key issues for such algorithms and present four new approaches developed for assessing conceptual similarity for concepts in concept maps. Two use precomputed summaries of structural and correlational information to determine the relevance of stored concepts to selected concepts in a new concept map, and two use information about the context in which the selected concept appears. We close by discussing their tradeoffs and their relationships to research in areas such as information retrieval and analogical reasoning.

See http://www.cs.indiana.edu/~leake/INDEX.html for additional publications in the Artificial Intelligence/Cognitive Science report and reprint archive maintained by David Leake.