OBOME - Ontology based opinion mining in UBIPOL

Abstract

Ontologies have a special role in the UBIPOL system, they help to structure the policy related context, provide conceptualization for policy domain and use in the opinion mining process. In this work we presented a system called Ontology Based Opinion Mining Engine (OBOME) for analyzing a domain-specific opinion corpus by first assisting the user with the creation of a domain ontology from the corpus. We determined the polarity of opinion on the various domain aspects. In the former step, the policy domain aspect has are identified (namely which policy category is represented by the concept). This identification is supported by the policy modelling ontology, which describe the most important policy – related classes and structure. Then the most informative documents from the corpus are extracted and asked the user to create a set of aspects and related keywords using these documents. In the latter step, we used the corpus specific ontology to model the domain and extracted aspect-polarity associations using grammatical dependencies between words. Later, summarized results are shown to the user to analyze and store. Finally, in an offline process policy modeling ontology is updated.