An Ontology Term Extracting Method Based on Latent Dirichlet Allocation

Ontology plays an important part on Semantic Web, Information Retrieval, and Intelligent Information Integration etc. Ontology learning gets widely studied due to many problems in totally manual ontology construction. Term extraction influences many respects of ontology learning as it´s the basis of ontology learning hierarchical structure. This paper mines topics of the corpus based on Latent Dirichlet Allocation (LDA) which uses Variational Inference and Expectation-Maximization (EM) Algorithm to estimate model parameters. With the help of irrelevant vocabulary, the paper provides better experimental results which show that the distribution of topics on terms reveals latent semantic features of the corpus and relevance among words.