A Semantic Approach to Extract the Final Answer in SBUQA Question Answering System

Abstract

In this paper we introduce the semantic approach of the answer extraction component of a question answering system called SBUQA. The answer extraction component gets the retrieved documents from a search engine according to a query (here NL question) as input. Then it represents both the question and candidate sentences using Lexical Functional Grammar (LFG), a meaning based grammar that analyses sentences in a deeper level than syntactic parsing. Then the meaning of the question and the candidate sentence will be expanded according to lexico-conceptual relations in a lexical ontology such as WordNet. The results of these expansions- stored in f-structure patterns- will be compared together exploiting our proposed extended unification algorithm. The answers will be ranked according to the type of found matches between the expanded meaning of the question and the expanded meaning of the candidate sentence (which contains the answer). There are four introduced types of matches from exact to different levels of approximate matching. After selecting the best (most matched) sentences, the final answer phrase will be extracted according to some predefined templates and ranked in descending order of relevance.