Morpheme and semantic based approach for Tamil question answering system

Abstract

Answering natural language question using the information from relevant document of a collection of documents or from relevant pages of the web is the major requirement of many users. User need a system that should be able to retrieve concise answers to questions posed in natural language. Design of such a system based on search engine is an important topic of research in the natural language processing field and much work has been done by many researchers in last three years. This system is named as Question Answering (QA) system. The research work of this theses is focus on solving a factoid type and definitional who type QA problem for Tamil. The design of QA System for Tamil is a challenging task due to the partially free word-order nature and morphologically richness of the language and no capitalization in proper noun. This thesis introduces various techniques for each phase of QA system process for natural languages with the characteristics of Tamil. Part of speech tagging, chunking of syntactic structure and named entity recognition are essential preprocessing works to make ease for deep processing of any languages with the characteristics of Tamil. The information extraction part deals with extraction of information for factoid type questions and for definitional type questions separately. The information extraction for factoid type question involves the extraction of related sentences, the ordering of relevant sentences by ranking based on relevancy and the extraction of semantically related information from the top five ranked relevant sentences. This research work provides techniques, procedures and algorithms in information retrieval, information extraction and answer generation components to develop a web based QA system for morphologically rich and partially free word order language using morpheme features and semantic information. newline newline newline