Personalized recommender system for technology enhanced learning using learners' metacognitive activities

Learning, an active cognitive activity, differs from one learner to another. This therefore suggests the need for personalized learning. Recommender systems in this context can be seen as a resourceful tool to provide appropriate learning materials that are tailored to the (personalized) learning needs and goals of the learner; and also to enhance learning. The development of personalized recommender systems typically involves a \textit{learner model} component, which is used to capture and store the personal information, preferences and other characteristics of the learner. While reading, learners engage in number of metacognitive activities e.g. text marking/creating highlights. These metacognitive interactions could serve as useful information for the learner model, to achieve personalization. In addition, the use of a probabilistic topic modeling based document retrieval (Latent Dirichlet Indexing) method makes it possible to provide finer grained multiple but topically related documents to facilitate learning. The current study investigates the effectiveness of using the highlights (a metacognitive activity) a learner makes while reading, as a preference elicitation method for the learner model. It also investigates the use of the Latent Dirichlet Indexing model to provide relevant recommendation of textual learning materials that enhance the personalized learning experiences of learners in a task-oriented activity. The experimental design allows the comparison of the performance, learner experience, learner interaction, and a number of other subjective analysis measures among two groups conditions; where one group receives recommendations based on the proposed methodology, and the second group receive random recommendations. The recommender system is integrated with nStudy, an online learning platform that provides a number of annotation tools (e.g. highlighting, tags) that support metacognitive activities. Findings show that the highlights learners create while reading serve as an appropriate input mechanism to guide personalized learning recommendations. Specifically, there was a significant difference in the learners' evaluation of the recommendation quality and accuracy between the two group conditions. The findings revealed that the learners in the experimental had positive perception of the recommendation quality and accuracy, which is also correlated to the user experience, and interaction.