Creating Multi-level Class Hierarchy for Question Classification with NP analysis and WordNet

Question answering systems provide answer to the questions using various question processing techniques and extracts answers from a set of documents. Finding the exact answer is more interesting and useful than getting a list of documents to look through and find the answer manually. Question answering is not same as a traditional document retrieval search engine where a set of relevant documents are returned in response of the query, whereas, in the question answering systems the response of the query is a concise and exact answer to the question. Typically, Question Classification (QC) is the first step in a Question Answering (QA) system. This phase is responsible for finding out the type of the expected answer by pruning out the extra information that is not relevant to extract the answer. Almost all the previous QC algorithms evaluated their work on the basis of a common class hierarchy already defined. The coarse grained classes Location, Entity and Numeric in the existing hierarchy have a fine grained class Other. We present the framework to create new fine grained classes to replace the Other classes. We also discuss the motivation behind the replacement and how the new fine grained classes may support the answer extraction. Additionally, we also present an automatic hierarchy creation method to add new class nodes using WordNet and Noun phrase parsing.