SEMONTOQA: A Semantic Understanding-Based Ontological Framework for Factoid Question Answering

Abstract

This paper presents an outline of an Ontological and Se- mantic understanding-based model (SEMONTOQA) for an open-domain factoid Question Answering (QA) system. The outlined model analyses unstructured English natural lan- guage texts to a vast extent and represents the inherent con- tents in an ontological manner. The model locates and ex- tracts useful information from the text for various question types and builds a semantically rich knowledge-base that is capable of answering different categories of factoid ques- tions. The system model converts the unstructured texts into a minimalistic, labelled, directed graph that we call a Syntactic Sentence Graph (SSG). An Automatic Text In- terpreter using a set of pre-learnt Text Interpretation Sub- graphs and patterns tries to understand the contents of the SSG in a semantic way. The system proposes a new fea- ture and action based Cognitive Entity-Relationship Net- work designed to extend the text understanding process to an in-depth level. Application of supervised learning allows the system to gradually grow its capability to understand the text in a more fruitful manner. The system incorpo- rates an effective Text Inference Engine which takes the re- sponsibility of inferring the text contents and isolating enti- ties, their features, actions, objects, associated contexts and other properties, required for answering questions. A similar understanding-based question processing module interprets the user’s need in a semantic way. An Ontological Mapping Module, with the help of a set of pre-defined strategies de- signed for different classes of questions, is able to perform a mapping between a question’s ontology with the set of ontologies stored in the background knowledge-base. Em- pirical verification is performed to show the usability of the proposed model. The results achieved show that, this model can be used effectively as a semantic understanding based alternative QA system.