A top‐down enriching approach for ontology learning from text

Abstract

SummaryTo allow better communications between computers and people, ontologies have been adopted in several application domains (web, medicine, industry, etc.). Ontology building exhibits a structural and logical complexity. To the end of making high quality domain ontologies, effective and usable methodologies are needed to facilitate their building process. In this article, we propose to extend the classical methods of ontology construction to design semantically richer ontologies. The objective of this article is to study the relevance of the latent Dirichlet allocation model that generates probabilistic topic models for each enrichment proposal by adopting a domain independent core ontology model. The fitted model can be used to estimate the similarity between documents as well as between a set of specified words/terms using an additional layer of latent variables which are referred to as topics. Experiments were conducted to measure the quality of our proposal against other solutions. Obtained results discussed here are satisfactory.