Research On Automatic Examination Generating System Based On Ontology

As the booming of Internet, information technology has brought a lot of new fresh bloods into different industries. As a result, tutoring in college gets a huge revolution in its method, content and mode. The development of information technology enables us not only manually process information, but also automatically and intelligently. Examination, as regular, is an important way to determine how well student learned and teacher taught. Traditionally, examination is always worked out on papers. Teachers, from their understanding, as the examination producer, organize and describe the tested knowledge in different ways. Currently, the generation of examination is always through manual generation and semi-automatic way. However, due to the complexity of Chinese sentence and knowledge description, the strategy of question generation existed so far has some shortages. For example, the shortages can be a lot of manual intervention and inefficient etc. In order to solve these problems, this paper proposes a new way to build automatic examination generating system, which based on domain knowledge base ontology. To do so, it will enable computer really understands such intelligent knowledge and produce a new strategy of question generation. As a result, it can automatically generate the course-specific examination, and at a great degree improve the efficiency of examination generation.As a core technique, ontology is used in this paper to build a domain knowledge base which formally describes domain concepts’layer, relations, definitions and features. Moreever, the paper generates the questions via sentence template of knowledge descriptions and the corresponding question templates. On the side of choice generation, the paper puts forward a concept similarity algorithm to compute the similarity of two concepts, and by which a concept-similarity-first tatic is proposed to generate the choices. Automatic generation of examination system, as a regular, contains two parts:items management (question automatic generation) and automatic questionnaire generation. The emphasis of this paper is on the former part—question automatic generation. The research is as followings:(1) The building of domain knowledge base on ontology. Domain knowledge base, as the knowledge resource of question generation, is comprised of concepts’definition, features, and relations etc in specific domain. To build such domain knowledge base, three steps are worked out:Firstly, the paper uses description logic to define concept sets, concept layers and concept relations sets; then based on concept relations, this paper defines abstract concepts and concepts’ features in axiom sets; finally, depicts the domain knowledge base by uniformly using ontology description language. As a consequence a domain knowledge base, which comprised of domain ontology, has been built.(2) A choice-question generation strategy which based on domain ontology knowledge base. The generation of choice-questions consists of two parts:items generation and choice generation. During generation of examination, we firstly retrieve the knowledge and get the description of them in domain knowledge base. Then, analyze the description’s construction and match the description pattern mode in order to automatically produce corresponding items. Meanwhile, according to the concept "family" tree formed in the domain knowledge base, the paper proposes a promotion on an algorithm, named concept similarity priority, to compute similarity of two concepts and as a return it can be used to measure the distract level of two concepts. After these former two steps a question repository is generated. Finally, we use a mutational standard generic algorithm to generate the questionnaire from the question repository. The experiment obtained shows the strategy proposed has these two advantages:the higher knowledge sharing and extension and higher distraction level of distracters. Besides, the paper simply describes how to generate fill-in-blanks and short-answer-questions. The description manily focuses on the generating algorithm and the generation of its corresponding answers.(3) Consider the premise description, we design and implement an ontology-based automatic examination generating system. Such system can well help teacher to generate corresponding domain-specific examination. The system adopts a layered framework J2EE to be its implementation. The system implemented has three layers:representation layer, business logic layer and data layer. And in representation layer, we use Java-server pages to represent the change of data between client and server. Whereas in business logic layer, this paper put Jena provided by HP Labs in use to query the domain knowledge stored in domain knowledge base. And data layer, as the name refered, provides data for the above service. The detailed function at each layer is described at the end of this paper. To layer the system, it can at greate degree improve the extension of this system, and a convenient maintainance of system.The automatic examination generating system this paper implements, well formally describes the domain concepts’layer, relations, definitions and features. As a consequence, it enables the system implemented have such features:well expression of knowledge, fortissimo ability of reasoning and extension. The system achieves the target that it can intelligently automatic generating questionnaire and well tests the knowledge’s emphasis, descriptions and so on as required. To build such system, it can efficiently figure out the problem of inefficient, slowly updating, more of labor-works and so on.