Course Ontology Based User's Interest Model Mining In E-learning System

With advanced communication strategy and abundant resources offered by modern information technology, E-Learning has become a novel educational mode and it can be used to deliver individualized, comprehensive, dynamic learning content to students timely, aiding the formation of knowledge communities, linking learners and practitioners with experts. It is those features that embody the active role that students play and accelerates the transformation from "teacher-centered" to "student-centered" education. This kind of E-Learning can supply the students with a personalized and self-adaptive learning environment. While designing and implementing this kind of E-Learning, one of the key techniques is to mine user's interest model precisely. And this field has attracted a lot of researchers.Considering the most important activities in learning process: self reading and interactive question answering activities, we present two approaches to capture user's interest about course content in E-Learning system. The main works in this thesis can be described as follows:(1)The definition of course ontology is given firstly. And then an example for course ontology, Artificial intelligence ontology, is constructed. The whole building process is described in detail. Furthermore, this course ontology has been used in our experiments.(2)We modified a text feature weighting method by adding term location information. We compare our method with traditional TF-IDF with AdaBoosting classifier in Weka3.5.7. Experiment results show our method leads to a higher classification performance in precision, recall and F-Measure.(3)An approach to mine user's interest form reading behaviors is proposed. The behaviors include actions such as underline, highlight, circle, annotation and bookmark while students are reading e-documents. A reading behavior record system named RBRS is implemented to record these behavior logs. Behavior table, weight matrix, behavior matrix and relative document matrix are defined and student's interest mining process is explained in detail. At last, an algorithm for mining user's interest from reading behaviors is presented. To prove our approach, we design an experiment with RBRS collecting experimental data. According to our algorithm, we compute each student's interest topics and the corresponding interest degree. Students' feedbacks are employed to evaluate the experiment results. It shows the feasibility of our approach.(4)The second approach we proposed is to mine student's interest model from interactive behaviors accumulated by interactive question answering system. An interactive Question Answering system named AIQA is developed. Students can pose a question, browse questions and answers and answer other questions freely in their favorite boards. All theseinteractive data can be used to construct each student's QASet. The mapping relations betweencourse ontology and QASet is constructed, which can be used to compute each user'sinterest about course content. The process of mining user's interest is presented in detail. At last, an algorithm for mining user's interest from interactive behaviors is presented. To prove our second approach, we design an experiment with AIQA collecting experimental data. According to the algorithm in this section, we compute each student's interest topics and the corresponding interest degree about the course content. At last we introduced students' recommendations acceptance rate to evaluate the experiment results. The evaluation results during 60 days show our second approach can mine student's interest model precisely and adaptively.