Terminological ontology learning based on LDA

Abstract

Ontology has extensive application in many fields, such as retrieval, information extraction and artificial intelligence et al. In this paper we describe a new approach about automatic learning terminological ontologies. this method make use fo the LDA model as concepts and builds relationship such concepts to learn ontologies. The method presents two measures, CP measure and L1 norm measure respectively, of computing semantic similarity between topics to organize these topics into hierarchy structure and forms the new ontology. Moreover, we design a method to determine the size of new ontology that is automatically created from text corpora, which can quantify the quality of the learned ontology in a natural manner. We evaluate our approach through GENIA corpus which is a text collections of biomedical literature. And the experiment results demonstrate the validity and efficiency of proposed method.