Content Recommendation Based on Education-Contextualized Browsing Events for Web-Based Personalized Learning
Wang, Feng-Hsu, Educational Technology & Society
The WWW contains a huge amount of learning content and is now in widespread use for delivering on-line learning content in many large-scale education settings. However, the huge amount of learning content has presented a problem to on-line students. Students are apt to get lost in the huge content space. Therefore, personalized learning becomes an important mechanism of a learning system that can guide students by automatically recommending learning content to their needs in a just-in-time manner (Zaiane, 2001).
The target of personalized learning varies with the types of learning needs it is designed to (McNaught, Kennedy & Majoor, 2002). For example, for just-in-time training, the main focus is on delivering appropriate information that workers need to solve problems, perform specific tasks or update their knowledge and skills. To perform such a type of personalized learning, specific knowledge about the task structure is required, such as in AIMS (Aroyo & Dicheva, 2001). On the other hand, for Web-based learning where students build the domain knowledge by studying learning content and navigating through a rich set of learning resources, personalized recommendation focuses on providing the next-step browsing suggestions so that students can build knowledge effectively with no disorientation in the learning environment (Kinshuk & Lin, 2003). This type of personalized content recommendation is the main focus of this research.
As to the design of personalized content delivery platforms, a new eLearning application, called Learning Content Management System (LCMS) (Brennan, Funke & Anderson, 2001), has been developed as a critical component of the personalized learning paradigm in which the emphasis is shifted from the knowledge of the instructor to the knowledge inherent in the content. Besides, a significant change is taking place in the way on-line learning is going in recent years. This change is the advocacy of learning objects to support personal learning needs. The concept of learning objects features the reusability of learning resources in whatever contexts they can be applied (Mohan, 2004; Mphan & Greer, 2003). Based on the key idea of LCMS that it will deliver what is needed at the time when it is needed, personalized content recommendation has become a very important learning support in LCMS (Brusilovsky & Vassileva, 2003; Denaux, Dimitrova & Aroyo, 2004).
One approach to personalized content recommendation is knowledge-oriented, which is based on three types of knowledge: domain ontology, content knowledge and student models (Papanikolaou & Grigoriadou, 2003; Sampson, Karagiannidis & Cardinali, 2002; Mittal, Krishnan & Altman, 2006). Domain ontology plays a shared language base for representing content knowledge and student models. A knowledge-based recommender could give highly individualized content recommendations. However, this approach incurs the cost of developing knowledge bases and is limited by the extent to which the student model is accurate.
Another approach to personalized content recommendation is to acquire knowledge about effective learning experiences and then share the knowledge with future users (Mulvenna, Anand & Buchner, 2000). As far as content recommendation is concerned, the ability of a web-based learning system to keep traces of students' browsing behavior can be exploited to promote the learning paradigm of learning by other's learning experiences (Najjar, Wolper & Duval, 2006). For those online courses that operate in a yearly cycle, which are common in most universities, the browsing traces can be accumulated and enriched year by year. With an appropriate analysis of these traces, navigation patterns can be discovered and shared with future students (Liang & Leifer, 2000; Najjar, Wolper & Duval, 2006).
There have been several research efforts to clarify the relationship among student cognitive characteristics, navigation patterns and learning performance. …