Recommender system applications

26 Chapter 2: Related Works

Tab.2.1 The main commercial recommender systems

Field Systems

E-Commerce Amazon.com, eBay, Levis, Ski-euripe.com

News GroupLens, PHOAKS, P-Tango

Music CDNOW, Ringo, CoCoA

Movie MovieLens, Netfilx.com, Moviefinder.com, Reel.com

Web Page Fab, Foxtrot, ifWeb, MEMOIR, METIOREW, ProfBuilder, QuIC,

Quickstep, R2P, Siteseer, SurfLen

represented by the order in which recommendations are given. The simplest form of a suggestion is the recommendation of a single item. In a list, the best items are at the top. Relevance can also be visualized using e.g. different colors and font sizes, or shown via ratings. Ratings can use different scales and different symbols such as numbers or stars.

The designer must take all of these factors into account in the early conception of the system. The following session, we turn our focus on the applications of recommender systems.

Chapter 2: Related Works 27

products they would like to purchase [Yong et al., 2005]. A number of recommendation systems have been proposed for different businesses. (Tab.2.1)

By using these recommender systems, users easily determine which products to buy or to select.

Products can be recommended based on the top overall sellers on a site, on the demographics of the user, or on an analysis of the past buying or selecting behavior of the user as a prediction for future buying or selecting behavior. The forms of recommendation include suggesting products to the user, providing personalized product information, summarizing community opinion, and providing community critiques.

2) Recommender systems for e-learning

Recently, in the e-learning area new methods and tools have been developed in order to improve and

Tab.2.2 Recently recommender systems for e-learning

Authors Scope Approach

Paramythis et al., 2004

personalized course delivery addressing students’ individual

needs e-learning standards and adaptive Zaïane, 2002, Andronico

et al., 2003 recommend learning activities agent-based, association rules

Hsu, 2008 reading lessons association rules

Yang et al.,2005 E-Learning resource

recommendation community-based, group agent Farzan and Brusilovsky,

2006 annotated courses case-based, community-based

Godoy and Amandi, 2008 collaborative tagging Content-based, social tagging Tang and McCalla, 2005,

2009. Tang, 2008 papers on the Web collaborative filtering, pedagogical value

Drachsler, H., Hummel,

H., & Koper, R.. 2008 RS for lifelong learning Learning network, content-based, collaborative filtering

Chen et al., 2004 courseware recommendation Fuzzy Item Response Theory Mohammad et al., 2008 E-Learning resource

recommendation Fuzzy-genetic, hybrid

Yu, et al., 2007 learning materials context-adaptable, context-aware, ontology-based

Peis et al., 2008 analysis of the state of the topic semantic web, ontology

Luis et al., 2008 learning materials Bayesian network, Fuzzy Set Theory Wan et al.,2007, 2008 e-NOTEBOOK, Collaborative

Notebook Hybrid filtering, implicit rating, time

28 Chapter 2: Related Works

customize the e-learning web sites according to learners’ necessities and tastes. Information technologies have an important role for CSCW (Computer-supported cooperative work). Digital libraries, corporative portals and Intranets, Web chats, newsgroups and e-mail listings allow people to share knowledge directly or indirectly. The most successful tool in these fields has been the recommender system. The aim of making use of the recommender systems in e-learning is to assist people to find the best alternatives that satisfy their necessities using recommendations, leading them to interesting items, or hiding those useless and unattractive ones.

Recommender systems are becoming an important way to support knowledge acquisition; various recommender systems have been developed (Tab.2.2). Brusilovsky discusses applications of adaptive hypermedia systems (a kind of recommender system) in educational environments, to support students in the learning process [Brusilovsky, 1996]. Most e-learning systems have made efforts in the promotion of learning efficiency of new concepts in a course. Wang et al. [Wang and Tsai, 2009] propose the review course composition system with adopts the discrete particle swarm optimization to pick the suitable materials, and can be customized in accordance with the learner’s intention. Huang et al. combined with some auxiliary materials like Blogs to assist the learner. If the exam result is not qualified, the systems will recommend some auxiliary materials associated with current subjects instead of original teaching materials [Huang et al., 2008]. Tang and McCalla propose a paper recommender systems in e-learning domain which consider pedagogical factors, such as paper’s overall popularity and learner background knowledge factors that are less important in commercial book or movie recommender systems [Tang and McCalla, 2005, 2009] [Tang, 2008].

A personal recommender system in learning networks in order to provide learners advice on the suitable learning activities to follow is proposed [Drachsler et al., 2008] [Drachsler, 2009]. And their system combines memory-based recommendation techniques that appear suitable to realize personalized recommendation on learning activities in the context of e-learning. Hsu develop a system based on association rules to recommend useful English learning materials in an ESL [Hsu, 2008].

Recommender systems are the technical response to the fact that we frequently rely on

Chapter 2: Related Works 29

recommendations when confronted with decisions in a field where we have little or no knowledge. It is a recent development to see the process of navigation in the Internet not as an isolated procedure of a single user, but to make the net knowledge of individual users available to others.

3) Moving from e-commerce to e-learning recommender systems

[Benyon 1993] emphasized that it is important to address many factors, such as user’s status, expertise, preferences and etc., in an appropriate way to enhance the usability and satisfaction for using the software. Since recommender systems are strongly domain dependent and it is therefore not always possible to apply one recommender system from a domain with a specific recommendation purpose into another domain with different domain characteristics [Adomavicius and Tuzhilin, 2005]. For example, the most famous Amazon.com’s recommendation algorithm will hardly be applied for recommending learning resources to a learner, because they require a deeper reasoning.

Recommender systems provide consumers with information to help them to decide which products to purchase, and they enhance e-commerce sales in three ways: (i) Converting browsers into buyers (ii) Increasing Cross-sell (iii) Building loyalty [Schafer et al., 1999]. On the other side, compare with e-commerce, recommender systems in e-learning enhance learning as below three ways [Drachsler et al., 2007] [Drachsler, 2009]: (i) they have as main recommendation goal to support learners with learning resources to help them to achieve each learning goal of themselves. Hence, considering how the certain competence level the learner has been mastered is necessary. (ii) Regarding the

“Increasing cross-sell” goal, recommender system in e-learning surely needs to suggest additional learning activities to learners based on those learning goals they aim for. (iii) Learners are satisfied if they get suitable recommendations for their specific learning goals.

In particular in the field of recommender system for e-learning, there is a need for more flexibility and control on the recommendation process. The relevance of courses or course elements, and the order, in which they could or should be followed, depends on many factors. To name a few:

• The learners’ knowledge, interests, goals and tasks, background, individual traits, context of

30 Chapter 2: Related Works

work should be considered [Brusilovsky and Mill´an , 2007].

• Recommender systems in e-learning are to support the learners in their competence development in order to achieve a specific learning goal [Drachsler et al., 2008] .

• Learner modeling has to use information about the learning process, which is closely connected to guidelines from educational, psychological, social, and cognitive sciences [Aroyo, 2006].

Whereas most current tools for e-learning provide relatively sophisticated functionality for the personalized learning environments, only little support is provided on the understanding of learner, learner relationships. In addition, recommendation for learners only considers the learner’s preference, and ignores learner’s competence. Therefore, in this thesis, we will consider these problems and propose a new approach for e-learning recommender system. The following two sections present related works on methodologies utilized in proposed recommendation approaches.