A SEMANTIC REPRESENTATION FOR IMPROVING THE PERFORMANCE OF COMPARATIVE QUESTION ANSWERING SYSTEMS

Abstract

Question Answering (QA) is considered a vital research area in Information Retrieval (IR) and Artificial Intelligence (AI). Most QA models depend on bag-of-words as a representation of the documents and questions when answering more than just the semantic representation in relevance estimation of information which in turn reduces the effectiveness and performance of the QA systems. The aim of improving QA systems is to increase accurate answers and provide access to a large number of related passages. Overall, the QA system is incapable of understanding the natural language with high accuracy. This could be attributed to its richness of hidden and embedded information. Accordingly, the main aim of this research is to design a model that uses the semantic approach to search and retrieve relevant answers in the QA system, particularly for comparative questions. This research proposes a new model which will improve question classification and document representations by using semantic analyses. The classification of these questions can be improved with an additional classification layer that can be added in the question processing stage for further semantic classification. This layer will be responsible for grouping the question as comparative or noncomparative.