Practical approach to knowledge-based question answering with natural language understanding and advanced reasoning

• Main Author: Wilson, Wong Yik Sen
• Format: Thesis
• Language: English; English
• Published: 2005
• Subjects: Q Science (General); QA Mathematics
• Online Access: http://eprints.utem.edu.my/id/eprint/14849/; http://library.utem.edu.my:8000/elmu/index.jsp?module=webopac-d&action=fullDisplayRetriever.jsp&szMaterialNo=0000014667

The complexity of natural language and the open-domain nature of the World Wide Web have caused modem-day question answering systems to rely only on information retrieval techniques and shallow natural language processing tasks. This approach has brought about serious drawbacks namely restriction on the nature of question and response. This restriction constitutes the first problem addressed by this research. Through recent academic works, many researchers have begun to acknowledge the problem and agreed that the solution comes in the form of a new approach based on natural language understanding and reasoning in a knowledge-based environment. Due to the infancy stage of this new approach and practical consideration, the current practices vary greatly and are mostly based on only low-level natural language understanding, minimalist representation formalism and conventional reasoning approach without advanced features. As a result, not only were these systems found to be inadequate to solve the first problem but have also created the second problem, that is the limitation to scale across domains and to real-life natural language text. This research hypothesized that a practical approach in the form of a solution framework which combines full-discourse natural language understanding, powerful representation formalism capable of exploiting ontological information and reasoning approach with advanced features, will solve both the first and second problem without compromising practicality factors.The solution framework is implemented as a system called "Natural Language Understanding and Reasoning for Intelligence" (NaLURI). More importantly, two evaluations and their results are presented to demonstrate that the inclusion of more demanding features into a question answering system will not only allow for a wider range of questions and better response quality, but does not affect the response time, hence approving the hypothesis of this research.