Automatic domain ontology learning based on web mining
Weihong Liu · Journal of Tsinghua University(Science and Technology) · 2005
This paper proposes a web-based learning model to acquire domain ontologies and to quantify the confidence of concept relations. An ontology backbone was captured with an extensible pattern set and a distributional semantic model to find the general relations between concepts using association rules, as well as to prune and merge candidate ontologies. The confidence of concept relations was determined by the confidence of patterns, semantic distances and association features of concepts. Model parameters and thresholds were optimized with a recursive procedure of analysis - ontology learning - text enrichment. An experiment proved effectiveness of this model. This model solves the problem of lexicon or core ontology dependencies in traditional ontology learning methods.