An integrated approach for concept learning and relation extraction

Abstract

Concept learning and hierarchical relations extraction are core tasks of ontology automatic construction. In the current research, the two tasks are carried out separately, which separates the natural association between them. This paper proposes an integrated approach to do the two tasks together. The attribute values of concepts are used to evaluate the extracted hierarchical relations. On the other hand, the extracted hierarchical relations are used to expand and evaluate the attribute values of concepts. Since the interaction is based on the inaccurate result that extracted automatically, we introduce the weight of intermediate results of both tasks into the iteration to ensure the accuracy of results. Experiments have been carried out to compare the integrated approach with the separated ones for concept learning and hierarchical relations. Our experiments show performance improvements in both tasks. © 2009 Springer Berlin Heidelberg.