Academic journal article Educational Technology & Society

Utilizing Learners' Negative Ratings in Semantic Content-Based Recommender System for E-Learning Forum

Academic journal article Educational Technology & Society

Utilizing Learners' Negative Ratings in Semantic Content-Based Recommender System for E-Learning Forum

Article excerpt

Introduction

The emergence of web 2.0 has led to a revolution in education industry where interactive e-learning systems show fast and significant growth world-wide. Nowadays, interactive e-learning environments utilize online discussion forums as a medium for collaborative learning that supports knowledge sharing and information exchanging between learners that have different knowledge levels. Shana (2009) asserted that using online discussion forums as an instructional tool improves the students' learning performance through providing better cognitive and exploratory learning. Furthermore, the idea of utilizing online discussion forums for learning is strongly supported by Social Learning Theory (Bandura, 1977). However, the exponential growth of the available shared information on e-learning discussion forums, as well as the learners' limited time for studying have caused a difficulty for learners in discovering interesting information that is relevant to their learning context. To overcome this problem, we propose a novel e-learning recommender system that recommends interesting information to the learners, thus save learners' time and improve their learning performance.

Recommender systems are utilized for personalizing information sources for users by guiding them in a personalized way to interesting items selected from myriad of available options (Lops et al., 2011). Basically, recommender systems are classified into several categories based on the adopted filtering approach (Adomavicius & Tuzhilin, 2005). Content-based, collaborative and hybrid filtering techniques are the most common filtering approaches used in recommender systems. In content-based recommender systems, recommended items are similar to the ones that the user preferred in the past. This approach is effective in filtering items of textual form where each item is represented as a set of keywords that describe it (Lops et al., 2011). Vector space model is considered as one of the widely used algorithms in content-based recommender systems where items in this model are represented as weighted vectors of keywords in the vector space (Turney & Pantel, 2010). The similarity values between items are obtained based on the cosine of the angle between their weighted vectors. In contrast, collaborative recommender systems recommend items based on item's profile, where the recommended items are similar to the ones that have been preferred by similar users. Nearest neighbor algorithm is one of the most popular methods used in collaborative filtering (Ekstrand et al., 2011). It was first proposed in GroupLens recommender system by Resnick et al. (1994) to filter news articles for users. On the other hand, hybrid recommender systems were emerged to overcome certain limitations in both of content-based and collaborative filtering techniques by combining them using several ways (Adomavicius & Tuzhilin, 2005). One way to build a hybrid recommender system is by implementing both methods separately and combining their final predictions, while another way can be done by incorporating some characteristics from one filtering approach into another (Burke, 2002).

Unlike recommendations in other domains, recommender systems in e-learning domain should assist learners in constructing their knowledge in a contextualized progressive way rather than acquiring it. Constructivism Learning Theory states that learning is an active and contextualized process of constructing knowledge; thus, it emphasizes the importance of the active involvement of learners in constructing their knowledge in a contextualized progressive way (Fosnot & Perry, 1996). This theory strongly supports the idea of avoiding the information that is relevant to the learner's previous understandings (i.e., avoiding the negatively rated items), and keeping the recommendations relatively progressive to promote a contextualized and smooth learning process within a given framework or structure. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.