Improvement of Attribute Correlation Method and Proposal of Collaborative Attribute Method in Text Recommender Systems for E-Learners

E-learning is used in various places. However, many systems do not show advantages, such as online exams, and simply enumerate the teaching material, etc. In our An Individual Reviewing System (abbreviated AIRS), contents of each user are optimized according to recommendations using Collaborative Filtering (what we call CF). This system multiplies the load to the user by smoothly improving study efficiency. However, this CF method has disadvantages in that if insufficient data is available, recommendations may show poor accuracy. This is what we call Cold-Start problem. In this paper, to solve this Cold-Start problem, firstly we provided a solution of Attribute Correlation Method that uses metadata which are belonged to users. And, we experimented with this Attribute Correlation Method, but the good results were not obtained. Secondly, in order to improve this Attribute Correlation Method, we proposed a new approach (called Collaborative Attribute Method) is to address this Cold-Start problem and showed the experimental results.