Hierarchical Taxonomy Based Topic Recommendation in Informal Learning
Mikko Vilenius · Institutional Repositories DataBase (IRDB) · 2012
In this research we present a novel and effective method of topic recommendation and group forming to support informal learning in online discussions, centred in the use of the so-called Web 2.0 applications, such as social media and discussion boards. While topic recommendation and group forming have been researched in a well constructed and clearly defined environment, the illconstructed and sporadic nature of informal learning poses new challenges for the typically used methods. Our method is especially suitable for encouraging active and knowledgeable people to join a discussion. The nature of informal learning is very different from traditional, formal learning. Formal learning usually uses well defined sources and happens in a certain space or environment, like in a school, university, or on a specific course. In terms of e-Learning this refers to, for example, Learning Management Systems where we have detailed information on the learners and clearly set goals and learning items. Informal learning on the other hand uses a wide variety of sources ? both onand offline ? such as books, television, direct person to person communication, the Internet, etc. and can happen in a wide variety of places and social environments. Furthermore, informal learning is by nature self-directed and self-centred, and is centred on collaborative and social activities. It is usually ill-constructed, i.e. there is no clear structure to the order learning happens in and it changes dynamically in irregular patterns. In terms of e-Learning this means that we cannot grasp the whole learning process, and cannot define clear learning goals. We can only give support to the learners by providing them with useful resources. In the course of this study we are especially interested in providing the learners with other learners to discuss with. We will recommend learner initiated discussions to people who are interested and knowledgeable on that specific topic. This means creating groups in which learners collaborate within an open time frame. Due to the more open ended and uncontrolled nature of these discussions conventional recommendation methods do not provide with accurate recommendation and are not as such suitable for multifaceted and noisy online discussions. Therefore we added a Hierarchical Topic Taxonomy to our recommendation model. The hierarchical taxonomy acts as a sort of “divide-and-conquer”algorithm increasing the accuracy of recommendation / group forming. We use two data sets from online discussions to evaluate our method’s recommendation accuracy in comparison to conventional methods. In addition we compared our method with existing methods by means of user questionnaire in order to evaluate the quality of the recommendation and its suitability for informal learning. Evaluation of our method shows an increase in topic recommendation precision in comparison to other, often-used methods. The results of our questionnaire show that the method is especially suitable for recommending knowledgeable people to discuss a given topic, which in turn indicates a better suitability for supporting informal learning. As a secondary result, our research also yielded some interesting insights on user participation patterns in online discussions.