Design And Implementation Of Question Answering System Based On The Understanding

With today's information explosion, it's very difficult to acquire information, which brings forward higher request of information retrieval. In order to meet the demand of obtaining information quickly and accurately, the research of question answering system began to rise. As an advanced form of information retrieval, the question answering system can answer users'questions posed in natural language using accurate and concise natural language. However, most of the current question answering systems feedback answers using only simple model matching, and thus can not make the robot really"understand"the questions but just remain on the surface. To solve this problem, the main focus of this thesis is the design and implementation of question answering system based on meaning comprehension.Question answering system can be divided into four sub-modules, that is, question understanding, information retrieval, answer generation and the meaning base called"KnownNet"in this thesis. The question understanding sub-module analyzes and comprehends users'understanding and questions expressed in natural language. According to the understanding of questions, the information retrieval sub-module can retrieve"KnownNet"and find the right answer. Then the answer generating sub-module generates information in natural language to answer users'questions, using the well designed natural language generating algorithm, combined with the feedback information from the information retrieval part. And"KnownNet", as the core sub-module, supports the whole question answering system. In this thesis, the meaning and the understanding of meaning are defined. Based on these definitions, a model of"KnownNet"which can understand the semantics from the bottom is designed and realized in detail by simulating the thinking mode of human. Different from the traditional knowledge base,"KnownNet"is based on"understanding". On the basis of the established"KnownNet", the learning and accumulating function and the memory function of"KnownNet"are also designed and implemented in this thesis by imitating human brain's functions of thinking and memory, which is the recurrence of the learning function of human. Therefore, through the human-computer interaction, the robot can acquire knowledge from outside, and enrich its own"KnownNet".In the final part of this thesis, an event instance is designed and"KnownNet"is constructed accordingly. Under the support of the constructed"KnownNet", the realization process of the question answering system is simulated, and the performance of the question answering system based on understanding is tested.