QUESTION ANSWERING SYSTEM FROM UNSTRUCTURED DOCUMENTS FOR BANGLA LANGUAGE

Abstract

This research describes the design and development of an intelligent question answering technique for Bangla language. Currently, an information retrieval (IR) system does not provide specific answer of a user’s question but extracts documents that are relevant to the need of a user. In this study, the intelligent question answering system is similar to information retrieval as users submit question for finding answers from the given source documents. The objective of this research is to develop a proper information or answer retrieval system rather than document or passage retrieval which IR does actually. Here, the rule based and the machine learning based approaches are implemented just like as the current research trend. For a given question, all of the approaches in this study extract keywords, lexical and semantic features from the input to find concise information or answer from the single or multiple text articles that is relevant to the question. For time and quantity related questions, the system gives specific answer otherwise the system retrieves relevant information. In the experimentation stage, the comparison between the rule based and the decision tree based approaches are made. Besides, an empirical analysis based on the other existing question answering systems has been done.