Research Of Question Answering System Based On Comprehensive Information

Intelligence Question answering system is a human computer interaction system that assists people to obtain information. It accepts natural language questions from, searches information to answer user questions in Knowledge Base and feedback the answers to user in a concise, accurate and natural way.Question answering system is an important research field in artificial intelligence. It is considered as an important advance in the development of Artificial Intelligence. Question answering system involves Artificial Intelligence, Knowledge Engineering, Natural Language Processing, Machine Learning, Linguistics and etc.There are massive topics in question answering research. And information is the core issue. It includes "How to understand the information requirements form user question?","How to understand the knowledge from knowledge base (knowledge is advanced information)?","How to represent and organize the knowledge in knowledge base?","How to decide the matching status between the information form user questions and the information from knowledge base "and "How to express the answers using nature language?"The key point for the performance of question answering system is weather the information is utilized adequately. Our researches focus on two problems about information in question answering system. The first problems are that the information is not comprehensively and deeply described and understood. It is manifested that the syntactic information and semantic information are appreciated while the pragmatic information is neglected. And the relation among syntactic information, semantic information and pragmatic information is not paid attention to. These affect the understanding of information. The second problem is that the research on the relations between different parts of question answering system is not enough. Especially the response information from user is not deeply researched.We first introduce comprehensive information theory the trinity of syntactic information, semantic information and pragmatic information. It provides new ideas and methods to solve previous problems.The idiographic work is as follows:First, based on the comprehensive Information theory, pragmatic information is introduced in question answering.Pragmatic information is not concerned in related researches on question answering system. In this paper based on comprehensive Information theory pragmatic information is defined and used in question answering system. Based on comprehensive Information theory and the first information conversion theory six pragmatic factors are defined:(1) subject of pragmatic information,(2) the language signs in question answering system,(3) Knowledge base,(4) user intention,(5) the question answering contest,(6) the effect of question answering. Pragmatic information leads question answering system understands the information demand of the user more deeply, and decide the satisfaction of the answers more accurately.Second, using Speech Act theory form Austin to analysis user’s goal and the pragmatic information in question answering.Speech act analyse the implication of utterances. And a two layers speech act classification criterion is designed to analyse the goal of the user. And it is also used to analyse the speech act of answers form knowledge base to decide whether answers meet the goal of users.Third, the phrase based syntactic information and speech act template based pragmatic information is used to fuse syntactic, semantic and pragmatic information.Only using speech act to analysis pragmatic information meets problems. The pragmatic information is not sufficient and accuracy. It leads the difficulties to understand the semantic content of sentences. In this paper the syntactic, semantic and pragmatic information are fused to excavate the relations among them. Ranking learning is used to select the collect answers with the comprehensive Information features. Experiments proved that the fusing of information improves the performance of question answering system.Forth, the Analects of Confucius question answering system is designed.This system has the following two characteristics. First the knowledge is extended from corpus to FAQ and structured knowledge. Question answering system should grasp the multiple types of knowledge, which helps answer the question answering system. Question answering system also needs to have the ability to accumulate knowledge in the question answering process. Second, increase the interactive answering system and users. In the question and answer process multiple rounds of user interaction and question answering system can provide richer information. Use the full context of access to information, especially information and user feedback to answer user questions.The experiments indicate that these methods improve the performance of question answering system. It proves that the ideas are on the right way and the methods are effective. However, it is very complex to represent information and understand questions. Hence our work is not perfect and need to continue to improve.