International Journal of Scientific & Engineering Research, Volume 3, Issue 4, April- 2012 1

ISSN 2229-5518

Personalizing M-learning Systems Using

Recommender Systems

Mahsa Baniardalan and Mehregan Mahdavi

AbstractIn mobile learning, there is no limitation for the number of learners; it is more attractive than traditiona l learning, more convenient and cheaper for updating, quicker in terms of broadcasting, less bulky and it is easier to transfer information. In addition, the studi es show that today, personalization is an important prerequisite for attracting learners in order to consider their requirements an d interests and to provide proper con- text for the learners. This can result in improving the quality of learning. In this paper, a comprehensive model is presente d which includes person- al, contextual, and environmental properties. Based on these parameters, the recommender system suggests educational materials in a proper manner such that the learner is attracted to the learning system. The system’s performance could be evaluated using questionnaires.

Index TermsContent, Context Aware, Mobile Learning, Personalization, Recommender Systems.


—————————— ——————————
T the present era of information technology, it is poss- ible to have a mobile form for every application, such as is learning. The category of mobile learning is a new con-
cept which has been introduced after electronic education has been developed in order that learning and training can be available in everywhere and every time and it can be eva- luated as well.
Mobile learning has been given various definitions including: Mobile learning can be considered as a combination of ―dis- tance education‖ and ―electronic‖ because the trainer and learner are separated by a distance and the learning is facili- tated by advanced technologies using electronic tools [1]. Ac- cording to another definition, mobile learning is an introduc- tion to ubiquitous learning in which the learner can access to his or her required information everywhere and every time [2].
As the whole, this method of learning can be considered as
everywhere and every time learning through mobile tools
such as portable computers and mobile phones, with no need
for physical connection to the network [1, 4].
Mobile learning can provide effective methods of learning, specifically educational courses for those who are not able, for any reason, to attend in full time classes. Mobile learning has been successful in Europe and Africa [3].


Mahsa Baniardalan, Department of Information Technology, International

Campus, University of Guilan, Iran, E-mail:
Mehregan Mahdavi, Department of Computer Science and Engineering, University of Guilan, Iran, PH-01316690274. E-mail: mahda-
In another definition, mobile learning is defined as any learning in which learner is not in a preplanned and stationary place and the learning is processed through mobile technology [5]. In other words, mobile learning does not depend on a spe- cific place and it is flexible in terms of place of learning and mobile technology. In fact, the mobility of learning is on focus [17]. Therefore, the importance of mobile learning is that it makes possible to access learning everywhere [6]. In addition, it is interactive and enjoyable and it is easily used to learn effi- ciently and more amusingly.
The previous research has shown that educational outcome is not the same for all people and it often fails in spite of spending much money and consuming a lot of time. Users and learners in these systems differ in terms of their previous knowledge, learning achievement rate, age group, skill and profession and personal intention to learn [7]. If mobile educa- tional system is designed based on the user properties, the quality of electronic education will be improved.
This research presents a simple but efficient model for per- sonalizing mobile learning in order to meet all needs of learn- ers.
This paper is organized as follows: Section 2 presents con-
tent analysis and the way the required content is developed. It
also presents existing models. An efficient model is introduced
in Section 3, which considers different aspects and priorities of
learners in order to provide learners with suitable educational
materials according to their needs. Finally, Section 4 concludes
the paper.


2.1 Recommender Systems

A recommender system in electronic education recommends

IJSER © 2012

International Journal of Scientific & Engineering Research Volume 3, Issue 4, April- 2012 2

ISSN 2229-5518

the optimal content to learners. In recent years, a lot of studies have been conducted on the process of guiding users in educa- tional systems [11, 13]. Generating useful recommendations to a group of users for items or products that might interest them is the main purpose of recommender systems. Some examples of operation of industry-strength recommender systems are suggesting books on Amazon, or movies on Netflix.
Most of the time, the so-called information overload problem makes a lot of difficulties for users to locate the right infor- mation at the right time. Recommender systems have come out as an important answer to it. It has been completely spread out in variety of applications, mostly in the form of intelligent virtual assistants in a variety of e-commerce do- mains. Recommender systems have demonstrated to be effec- tive at delivering the user a more intelligent and proactive information service. They use techniques borrowed from user profiling, information filtering and machine learning, in order to make proper recommendations for products or services based on users’ preferences and needs [14].
Generally two recommendation strategies are more impor-
tant. One of them is Content-based recommenders which rely
on rich content descriptions of items (products or services)
that are being recommended [15]. A content-based movie re-
commender which typically rely on information such genre,
actors, director, producer is an example of and matches these criteria against the learned preferences of users in order to select a set of promising movie recommendations. Clearly, the designet of such a content-based system should perform con-
siderable knowledge-engineering activities since the required domain knowledge may not be readily available or easily ex- tracted. The collaborative filtering (CF) recommendation strat- egy takes advantage of the fact that in real world we often look at our friends for recommendations. It does not need item knowledgeand instead uses user profiles which also include the past rating history or bahaviour of the users. It is also called social filtering. By using ratings history of a set of suita- ble users, it can generate useful recommendations for a target user. The set of users that are chosen by the system are chosen based on the similarity of the ratings history with the target user [14].

2.2 Context Definition

Context is defined as a collection of all information about a certain entity [10]. In mobile learning, the collection of in- formation is related to properties and situation of learners which can be collected explicitly (asking directly from the user) or implicitly (the system draws out what it needs secret- ly from the user) based on the behavior of users and suggests the best alternative to the learners [18].
In fact, any information related to improved interaction among learners, software and environment is called context [8]. The learners’ needs are often latent and mostly they can be identified from habit, morals and even from learners’ mo- tivation. To decrease the interaction challenges, the system designer must identify learner properties to know how the learner can be attracted to the system and how he or she can constantly attempt learning [9].

2.3 Personalization Process

Personalization is based on the modeling of learning con- text, in a collection of information that characterizes the situa- tion and environment in which the learner is. It comprises the information and assumption about the learner (such as per- sonal profile, goal, knowledge, interest preferences, interac- tion, and presentation history) and the information about the environment (such as location, devices, time, date and weath- er) [11].
It is clearly recognized that the personalization process is essential. Each person’s properties differ so his or her capabili- ties vary and we have to consider these properties. Personali- zation process is implemented using the identified context related to the learner. The purpose is to suggest a packet of activities to the learner based on his or her priorities and in- terests such as links, articles, materials, tests, homework and so on [10].
By personalization based on learning context, we mean
that certain learner-related information such as goal, situation, environment and assumptions are gathered and categorized as described in the followings.
Various definitions are previously provided for context. For instance, primary types of context can be considered as loca- tion, identity, time and activity [18]. Context can be considered in defferent categories. Computing context include informa- tion such as: network connectivity, communication costs, communication bandwidth. Information such as users’ profile and location create user context. There are also information on the environment that create environment context, such as noise levels, lighting situation, and temperature [19]. Moreo- ver, attributes of people and objects around a user and the changes in such information can be considered as context [20,
21]. Context-aware systems are defined as systems that are able to sense the context and act upon available context infor- mation [22]. In a deferent caregorization of context it can be considered as i) environment (physical and social), ii) self (de- vice state, physiological and cognitive), and iii) Activity (be- havior and task) [23]. In other words, context can be defined as any information that can characterize the situation of an entity (e.g. person, place, or object) [24, 25]. Context can also be cate- gorized as externally-imposed, externally-induced or internal- ly-induced [ 26]. An ontology-based context model may con- sidere time, place, user knowledge, user activity, user envi- ronment and device capacity [27]. In some previous research, a situation model was defined to give a view on the context model describing temporal properties. It used the following context parameters: variety, priority, granularity, implementa- tion, and cost-effectiveness [28]. Data gathering techniques about the context is an essential issue in pervasive computing [29]. Moreover, a prevous research in mobile learning has ar- gued that a context aware mobile learning system should also include the learner’s willingness to participate in the proposed learning activity [30].


In this model, as it is illustrated in Figure 1, assumptions and information about learners are categorized into many parts.

IJSER © 2012

International Journal of Scientific & Engineering Research Volume 3, Issue 4, April- 2012 3

ISSN 2229-5518

Figure 1. The proposed model for mobile learning recom- mender system
The recommender systems development process includes the followings:
1. Register: There are two modes. In the first mode, an in- tended user for learning, registers in his or her applica- tion and fills the questionnaire in the system and gives the primary information. In the second mode, the con- tent is sent according to the learner’s needs. Learner’s profile is empty and it is gradually filled implicitly in the system or partly filled by the learner in order for the system to identify the learner’s interests. The studies show that learner’s priorities are static. However, it is possible that the priorities change. We can permit the system to select automatically for the learner and over- come this challenge.
2. The learner’s application is sent to recommender system (of course, it can be spontaneous in certain cases and the information is sent to learner without primary applica- tion).
3. Content creation and collection: The information is col- lected in various ways such as the followings:
The learner defines certain variants such as de-
mographic data and level of education, favorite
orders such as using blue color for the texts or us-
ing audio or video. His or her field of interest
(math, science and so on), his or her goal of learn-
ing (practical use or professional use), physical
disabilities in terms of hearing, weight, social be- havior such as if the learner is introvert or extro- vert, whether the learner thinks positively or ne- gatively, what is the learner’s motivation to learn, what about his or her feelings and intelligence index, and so on.
Certain variants can be identified by the trainer
(learning theory).
Some variants can be identified by the parents
(expenditure limit and security measures). Variants which are identified by partnership
companies (evaluations).
Variants which others such as physicians, psy-
chologists, scholars, friends, producers and sup-
pliers in communication networks identify. Pretests which are given to learners to evaluate
and achieve certain variants (such as learning
style and learner’s level).
Some variants are achieved through software and hardware systems (such as health, weather, location of learner and maps).
Data mining techniques can develop learner re-
lated knowledge (such as interests, media ex-
change, equipment feature) or using online or
offline database such as social networks, state
Computer producers can monitor certain va- riants to build up data (such as current situation,
feelings and emotions, results and achievements). Tracking using sensor, GPS and wireless sensor network (WSN) or radio frequency identification
User personal properties and environment data can be ob- tained by these variants.
4. To request user’s properties from database with various formats such as text, audio, video, animation and so on which are collected and categorized by the experts for different groups.
5. To send the materials from database to recommender
system and through mediatory to the intended user.
6,7. After the user finishes the learning, the system ex-
amines the learner spontaneously by self test approach
and the result is stored in user personal profile in order
for the user know what should be sent during the next
session (if learner has learned well, more advanced ma- terials will be received and if it should be repeated, the previous material would be explained for the learner).
The compomemts of the proposed recommender system include the followings:
1. User (learner): The one who uses system and helps to improve it.
2. Mediator: The operational system which performs software and plays as an interface between the user and recommender system to facilitate their relations.
3. Recommender system: It contains three inputs and a collection of best recommendations for the user as the output. The total goal and system assessment and whether it is suitable or not will depend on the output.
4. Database: It contains related educational information for user personal profile and priorities which is de- signed and categorized by the experts and is presented to the user based on the user needs. Of course, for a real time and secure system, a high memory and high secure database is required. In addition, learning contents
must be also modified before being sent to the receiver’s

IJSER © 2012

International Journal of Scientific & Engineering Research Volume 3, Issue 4, April- 2012 4

ISSN 2229-5518

device by agents or filters that made the format trans- formation and serve contents adapted to them. This fil- ter should act depending on the type of devices [16]. The contents can be classified as: Text, Image, Audio and video
Text: Text can be easily converted to another mo-
bile device’s adapted format, but, due to the in-
crease on e-book readers, there are a lot of new
standards that modify the pagination and font
size automatically. Thesemake text easily reada-
ble on those devices. Among these new formats,
the most used are EPUB and PRC and LIT.

Image: Although most of the image formats can be watched properly on mobile devices, it is use-

ful but not essential, to scale them or to modify them before being sent.
Audio: Mobile devices, ipod/iphone and MP3 players support most of the audio formats, so a recording is not necessary. However, it will be necessary in case of video stream.
Video: For a good visualization, the video format must be transcoded [17] to be adapted to the basic mobile devices format as MP4 for iphones or ipods, and 3GP for rest [18].
5. User environment which includes:
Location: It is very important for applications, es-
pecially in location based learning, to use WSN,
RFID and GPS technologies.
Tools: Mobile tools play the main role for presen- tation and projection such as screen size, power of processing, data store, connection method, bat- tery consumption.
Time: The type of content depends on the in- tended time
Weather: The type and framework of content de- pends on the environment condition, crowded or not crowded, rainy or not rainy, cold or warm.
6. Learner’s properties: These properties may include: Personal profile: Personal information of learner
is used to classify the learner in a proper category and to assign him or her suitable learning con- tent.
Goal: Learners have a specific goal when they use
a learning system. By identifying their goals, the
system can supply the proper content for the
learner. So, it can be regarded as a filter which is
able to classify the learners with the same goal. A history of interactions and presentations: It is
useful in terms of: (i) The learner is aware of what he or she has learned ( for example, the learned books can be colored blue), and (ii) the personali- zation process can be improved by the system.
Knowledge level: It seems that it is the most im-
portant role of learning system and can be meas-
ured by scalable tests in terms of quality (in three
category: beginner, middle and advanced) and
quantity (0-10). The system does not have any in-
formation about the learner so it updates know-
ledge model constantly. For example, when a learner learned the material, the system examines and according to the results, it raises or decreases the knowledge level (knowledge propagation process).
Interests: They are categorized as well and if the
learner shows interest towards the content, it will
be explained in detail and if not, the content is
overviewed. It is noted that at first, the interest
rate is null and it is gradually improved and up-
Priorities: The learner can set up what is pre-
sented and how it is presented (page by page, as slides or all-in -one). The alternatives can be vari- ous media such as audio, video, animation, text and so on which can be presented by spontane- ous selection in the system or by directly selection of the learner.
Figures should de numbered consecutively as they ap-
pear in the text.


In general, learners can learn using various tools, in differ- ent environments and through many networks with diverse levels of learning and interests. Learner must be supported by adjustable recommender systems. The suggested model con- siders the entire user’s needs and tailors the content properly for the learner using the database of the system. But the prop- er feedback is implemented by post-tests. Although the pre- vious models considered certain useful details, in the present paper, we attempt to design a model which incorporates entire needs of the user and suggests the proper content in order to meet the needs of user and raise the satisfaction of learner.


[1] Safarzade, M. and Manuchehri, K., ―Learning Via Mobile Electronic Devices‖, 2nd International Conference Electronic Municipality, Te- hran, Irans Interior ministry, 2009.

[2] Yazdi, J. and Kahani, M., ―Resource Management in a Semantic Mo- bile e-Learning Grid‖, 12nd International Conference on Computer Society of Iran, Tehran, Iran, 2006.

[3] Abdolahi, M., ―Mobile learning and role of Future of E-learning‖,

2nd Conference on E-learning. Zahedan, Iran, 2007.

[4] Ataran, M., ―Curriculum prospective.‖Monthly School Tomorrow.

Five years.

[5] O’Malley, C., vavoula, G., Taylor, J., sharples, M., Lefrere, P., Lons- dale, P., Nismith, L. and Waycott, J., ―D4.l Guidelines for learning in a mobile environment‖, MOBllearn, UoB. OU, wp, d4.1/1.2. 2005.

[6] McFarlane, A., Roche, N. and Triggs, p., ―Mobile learning: research finding. Coventry‖, Becta. [online] vaila- ble: ments/research/mobile_learning_july07.pdf , 2007

[7] Jane, Y-K and Joy, M., ―A context-aware personalized m-learning

application based on m-learning preferences‖, International Journal

of Mobile Learning and Organisation, Vol. 5, No. 1, pp. 1-14, 2010.

[8] Jappinen, A., Nummela, J., Vainio, T. and Ahonen, M., ―Adaptive

Mobile Learning Systems-The Essential Questions fromthe Design

Perspective‖, In the Proceedings of MLearn2004, Roma, Italy, 2004.

IJSER © 2012

International Journal of Scientific & Engineering Research Volume 3, Issue 4, April- 2012 5

ISSN 2229-5518

[9] Yudelson, M., Gavrilova, T., and Brusilovsky, P., ―Towards user modeling meta-ontology‖, Proc.of 10thInternational User Modeling Conference, Springer Verlag, pp. 448-452, 2005.

[10] Economides, A., ―Adaptive context-aware pervasive and ubiquitous learning‖, International Journal of Technology Enhanced Learning, Vol. 1, No 3, pp. 169-192, 2009.

[11] Zhou, R., and Rechert, K., ―Personalization for location-based E-

learning‖, The Second International Conference on Next Generation Mobile Applications, Services, and Technologies, IEEE Coputer so- ciety Washingtone. DC, USA, pp. 247-253, 2008.

[12] Dunn, R., & Dunn, k., ―Teaching students through their individual

learning style: A practical approach‖, Reston.VA: prentice-Hall, 1978.

[13] Uday Bhaskar, N., and Govindarajulu, P., A design methodology for acceptability analyzer in context aware adaptive mobile learning sys- tems development. IJCSNS International Journal of Computer Science and Network Security, Vol. 8, No. 3, pp. 130-138, 2008.

[14] O’Donovan, J., Smyth, B., ―Trust in recommender system‖, in pro-

ceeding of 10th international conference on intelligent user interfaces, Acm press, California, USA, pp. 167-174, 2005.

[15] Melville, P., Mooney, R., and Nagarajan, R., ―Content boosted colla- borative filtering in improved recommendation‖, Proceedings of the Eighteenth National Conference on Artificial Intelligence(AAAI-

2002) , Edmonton, Canada, July 2002.

[16] Houser, C., and Thornton, P., ―Poodle: a course-management system

for mobile phones‖. Wireless and mobile technologies in education, Proceedings of the third IEEE International Workshop on Volume, Los Alamitos, California, IEEE Computer Society, pp. 215-215, No- vember 28-30, 2005.

[17] Reveiu, A., Dardala, M., Furtuna, F., ―Using content multimedia data

retrieval for multimedia content‖, Proceedings of the 12th Int. Conf on Human-Computer Interaction HCII, Vol.3 LNCS 4552: Lecture Note in Computer Science, pp. 486-492, 2007.

[18] Schilit, B.N., Theimer, M.M. and Welch, B.B. ―Customizing mobile

applications‖, Proceedings of the USENIX Symposium on Mobile

and Location-Independent Computing, (USENIX Association), pp.

129–138, 1993.

[19] Schilit, B.N., Adams, N. and Want, R., ―Context-aware computing applications‖, Proceedings of IEEE Workshop on Mobile Computing Systems and Applications, Santa Cruz, California, 1994.

[20] Becker, C. and Durr, F., ―On location models for ubiquitous compu-

ting‖, Personal Ubiquitous Computing, Vol. 9, Issue 1, pp. 20-31,


[21] Schilit, B.N. and Theimer, M.M., ―Disseminating active map informa-

tion to mobile hosts‖, IEEE Networks, Vol. 8, No. 5, pp. 22-32, 1994. [22] Ryan, N., Pascoe, J. and Morse, D. ―Enhanced reality fieldwork: The

context-aware archaeological assistant‖, Computer Applications in

Archaeology, British Archaeological Reports, Oxford, 1998.

[23] Schmidt, A., Beigl, M. and Gellersen, H.W., ―There is more to context

than location‖, Computers and Graphics Journal, Vol. 23, No. 6, pp.

893-902, 1999.

[24] Dey, A.K. and Abowd, G., ―Towards a better understanding of co n- text and context-awareness‖, Proceedings 2000 Conference on Hu- man Factors in Computing Systems (CHI), The Hague, The Nether- lands, 2000.

[25] Tamminen, S., Oulasvirta, Toiskallio, K. and Kankaninen, A., ―Un-

derstanding mobile contexts‖, Personal and Ubiquitous Computing, Vol. 8, Issue 2, pp. 135-143, 2004.

[26] Hill, J.R. ―Reflection on resource-based learning environments: Con- tinuing the exploration of opportunities and obstacles‖, International Journal of Knowledge and Learning, Vol. 3, No. 1, pp. 12-29, 2007.

[27] Bouzeghoub, A., Do, K.N. and Lecocq, C., ―A situation-based deli-

very of learning resources in pervasive learning‖, E.Duval, R. Kla m-

ma, and M. Wolpers (Eds.): EC-TEL 2007, LNCS 4573, pp. 450-456. Springer-Verlag Berlin Heidelberg, 2007.

[28] Bayoumi, F., ―Context aware systems: Present and future‖, Proceed- ings of the IASTED European Conference, Internet and Multimedia Systems and Applications, Chamonix, France, pp. 208-213, 2007.

[29] Roibas, A.C. ―Teaching appropriate ethnographic methodologies for pervasive computing‖, International Journal of Teaching and Case Studies, Vol. 1, No 1/2, pp. 135-145, 2007

[30] Bhaskar, N.U. and and Govindarajulu, P., ―A design methodology for acceptability analyzer in context aware adaptive mobile learning systems development‖, International Journal of Computer Science and Network Security, Vol. 8, No. 3, pp. 130-138, 2008.

IJSER © 2012