Unsupervised Multilingual Ontology Learning

Multilinguality poses a big challenge to the growth of semantic web. In order to develop multilingual applications we need to develop ontologies which can be shared across languages. In this paper, we propose an unsupervised learning algorithm to automatically learn multilingual ontology from unstructured text. We propose three different approaches for multilingual ontology learning, Dictionary based method, parallel corpus based method and Latent Dirichlet Allocation (LDA) based method. While the first two approaches require availability of dictionary and parallel corpus, the LDA based approach does not require any special resource. We have conducted our experiments for two languages, English and Hindi, however the proposed method can be adopted for other languages also.