A Reasoning Methodology for CW-Based Question Answering Systems

Abstract

Question Answering Systems or (QA systems for short) are regarded as the next generation of the current search engines. Instead of returning a list of relevant documents, QA systems find the direct answer to the query posed in natural language. The key difficulty in designing such systems is to perform reasoning on natural language knowledgebase. The theory of Computing with Words (CW), proposed by Zadeh, offers a mathematical tool to formally represent and reason with perceptive information. CW views a proposition in natural language as imposing a soft/hard constraint on an attribute and represents it in form of a generalized constraint. In this paper we develop a reasoning methodology for the restricted domain CW-based QA systems. This methodology takes, as input, the knowledgebase and the query in form of generalized constraints and organizes the knowledge related to the query in a new tree structure, referred to as a constraint propagation tree. The constraint propagation tree generates a plan to find the most relevant answer to the query and allows improving the answer by establishing an information-seeking dialog with user.