Active, Context-dependent, Data-centered Techniques For E-learning: A Case Study Of A Research Paper Recommender System
T. Tang & G. McCalla
CHAPTER 6 Active, context-dependent, data-centered techniques for e-learning: a case study of a research paper recommender system T. Tang1,2 & G. McCalla2 1Department of Computing, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong. 2ARIES Laboratory, Department of Computer Science, University of Saskatchewan, Saskatoon, Saskatchewan, Canada. Abstract In this chapter we discuss an e-learning system that recommends research papers to students wishing to learn an area of research. Recommender systems in the e-learning domain have specific requirements not present in other domains, most importantly the need to take into account pedagogical aspects of the learner and the system (such as the learning goals of each) and the need to recommend sequences of items in a pedagogically effective order. First, the architecture and some of the basic methodologies of the system are presented. Then, two studies are presented showing some of the ‘pedagogical’ characteristics affecting recommendations and comparing two recommendation algorithms: one that is content-based and the other that employs hybrid content-collaborative filtering and clustering techniques. We then generalize from the recommender system to discuss a general approach to the design of an e-learning application called the ecological approach, which is centered on finding patterns in learners’ interactions with learning objects and using these to actively compute information relevant to the application that is sensitive to the current end use context, especially to characteristics and goals of the learner. The ecological approach holds the promise of overcoming many problems in e-learning that have made systems controlling and unresponsive.