An NLP Based Question Answering System for Semantic Web

Abstract

The act of querying and finding large and heterogeneous content, such as those found on the Web, has become a more difficult undertaking as the semantic web has grown in popularity. To make the semantic web vision a reality, some user friendly interfaces are required to assist users in querying and searching the vast and heterogeneous information space. Because of the complexities of natural language, question answering systems on the semantic web present a variety of challenges and, as a result, research opportunities. newlineThe goal of this thesis is to create a question-answering system that will provide the most relevant response to the users query. Though there are many existing question answering systems and a literature survey was conducted to find out the gaps in these QA systems. Some of the important gaps included are The same query can be expressed in various ways, so question reformulation is required Existing frameworks use different techniques for knowledge base representation which are not efficient for large database There is a need to assign appropriate rank to the answer There is a need for real time question answering to make the system interactive Existing systems unable to generate appropriate answers for reasonable queries such as how and why. newline To address the aforementioned gaps, a novel semantic question answering system is proposed in order to provide the correct and timely answer to users queries from a knowledge base. We offer a semantic web based solution for question answering that leverages natural language processing to analyze and interpret the user query in this research work. To determine the importance of each response returned by the system, it uses a Total Answer Relevance Score. The results obtained are quite encouraging. To reduce the perceived latency of the user, the system will return the answer in real time from the SW repository if the user fires the same question again. newline newline