Research And Application Of Question Answering System Based On Course Knowledge

With the development of computer and network technology, modern distance education has been paid more and more attention, and changing the way of people's life and learning at amazing speed. Under the model of distance education, people can be free from the restriction of time and place, accepting education. It provides convenience to people, but also there are some problems. First, under the environment of network, there are a large number of students, and due to restrict to time and energy, the teacher impossible to answer every student's questions; Second, when student's questions are not answered quickly, it will be decreased the students'enthusiasm for autonomic learning; Last, the teacher have to answer repeated and similar questions. In order to solve the problems in distance education, the paper proposes based on course knowledge question answering system by research and analysis the knowledge and frequently asked questions in university course "computer operating system". Question answering system (QAS) allows describing questions in natural language by user, and then submitted to QAS, the QAS return a short and accurate answer. This is a new solution model based on natural language.The major research contents in paper are as follows:Firstly, in the stage of question analysis, reference to the question classification methods that has been proposed, combined with the characteristic of frequently asked questions in paper, proposes a question classification method based on question word and domain word and auxiliary word. Six question types are proposed according to the question classification method, they are definition type, reason type, relation type, list type, method type and other type.Secondly, in the stage of question similarity computing, reference to the question similarity computing methods that has been proposed, combined with the characteristic of questions in paper, proposed a similarity computing method that based on semantic of keywords. The method is fully considered of surface layer information of questions and combined some semantic information.Then, through analysis and statistics the questions in frequently-asked question library, finding every question belongs to a particular question type, therefore proposes a method based on question type to determine candidate question set. Experiments have confirmed that the proposed method reduced search space, and improved the system efficiency and accuracy.Lastly, Design and implementation of question answering system based on course knowledge. The paper has established a test set which contained 100 questions to test system performance, the result shows that the system's accuracy was 67%, verified the feasibility of the paper proposed method and question answering system has some practical.