Information Retrieval System using UNL for Multilingual Question Answering

Abstract

Efficient storage of information in a form which is independent of language forms the basis of enabling information sharing smoothly across various channels. Also the information retrieval systems build so far have the basic tendency to provide the user with the specific information that has been demanded by him. Refined query processing has facilitated easy retrieval across various regions where different languages are spoken using only a single corpus. This can be really helpful for designing a Question Answering system that does not require hard wiring of corpus. This requirement of having a single corpus for multiple languages is a challenging one and this thesis explains the solution using UNL. UNL is a widely known platform meant to receive,distribute and understand the information on multingual basis. It provides the computer systems with huge amount of knowledge that is accessible as well as understandable. In this thesis a very basic use case that is based on the semantic approach to handle the user based queries has been presented. Instead of returning whole bunch of related information to the user query , the system returns specific answer that gives the most descriptive infromation about the target term. This proposed system has been constructed to cater to the applications of the natural language processing and finally build a model that is language independent. The thesis describes the process of designing a UNL based QA system which is capable of handling the factoid based user queries such as what, where, what, why and which using a single language independent corpus. This system can service all the language provided their EnConverter and DeConverter are available. After testing the proposed system on the defined corpus it is seen that promising results have been achieved which thereby have proved the utilty of the system.