International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 1

ISSN 2229-5518

An Agent-Based Approach for Supporting

Ubiquitous Learning

Luiz Cláudio N. da Silva, Francisco M. Mendes Neto, Luiz J. Júnior, Raphael de C. Muniz

AbstractA mobile learning environment provides students with a teaching method that would be not possible to be performed in a conventional web-based course. The use of Learning Objects (LOs) in a standard way consists of an effective way to allow, among other features, content reuse and interoperability between different Learning Management Systems. However, a problem that occurs frequently is the unsuitability of the content to the context in which the student is inserted. A context -aware mobile learning support environment allows to solve this problem. Thus, this paper presents an agent-based approach to context-aware recommendation of LOs in order to enhance the teaching process in mobile learning.

Index TermsMobile learning, Learning objecs, Learning management Systems, Agent, Context-aware.

—————————— ——————————


ISTANCE Education (DE) is a mode of teaching and learning which has been growing quite some years ago. The developments of computer networks, the
improvement of the processing capability of personal computers and the advance of multimedia technologies, among other factors, contributed to the creation of this scenario. However, despite consisting of an effective teaching method, still presents some challenges, among which can be highlighted the need for a computerized support appropriate to the characteristics of each person. Through this support is possible to automate increasingly the process of learning, making the teacher a facilitator, ceasing to be the main source of information and passing to drive the learning process [1].
One of the ways to provide DE is through the use of mobile devices, this modality is known as mobile learn- ing. This way of providing education allows that students and teachers can take advantage of the resources offered by mobile technologies. One of these benefits is the possi- bility to access, view and provide content irrespective of time and from any location [2].
However, even with the benefits offered by mobile learning, the particular characteristics of each student should be considered, including the resources which they hold. This is necessary not only to provide content that meets the needs of students, but also to provide content in an appropriate way to the constraints of mobile devices since they have distinct and limited resources. In this context, arises the concept of context-aware environ- ments. This kind of environment fits to the user, consider- ing information provided by the selfsame user, beyond


Luiz Cláudio N. da Silva is currently pursuing masters degree program in

Computer Science, UFERSA, Brazil. E-mail:

Dr. Francisco M. Mendes Neto is Lecturer PhD in Computer Science,

UFERSA, Brazil. E-mail:

Luiz J. Júnior is currently researching in Technological Center of Software

Engineering, UFERSA, Brazil. E-mail:

Raphael de C. Muniz is currently researchng in Technological Center of

Software Engineering, UFERSA, Brazil. E-mail:
those captured dynamically from his interaction with computing devices [3].
For the purpose of developing context-aware envi- ronments in the learning domain, called Ubiquitous Learning Environments (ULE) [4], it is essential that edu- cational content be created in a standardized manner. Thus, it is possible that a Learning Management System (LMS) shows the contents properly and reuse content in different contexts and from different repositories. Fur- thermore, it is possible to combine different contents, which, in turn, improves the production process and, consequently, reduce its costs [5]. An effective way to standardize educational content is through the use of Learning Objects (LOs), which consist of small units of content that can be used, referenced and reused during a learning process [6], [7].
Given the relevance of the topic, the present work has as its general goal to provide a learning environment, through the use of mobile devices, to help and fit the needs of the students, according to characteristics of the context in which they are inserted.
This paper is divided into seven sections. Section 2 presents an overview of multiagent systems (MAS) to support learning. Section 3 describes the learning objects, as well as the standards used in their development. Sec- tion 4 presents the concepts and inherent characteristics in ubiquitous learning environments. Section 5 discusses related works. Section 6 describes the agent-based ap- proach proposed in this paper and the role of each agent. The last section presents our final remarks and a discus- sion of future works.


In despite of different definitions of agents that can be found in the literature, there is still no consensus on the issue. However, it is possible to construct a concept from the definitions given by researchers.
According to [8], agents are autonomous software enti-

IJSER © 2011

International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 2

ISSN 2229-5518

ties that perceive their environment through sensors and that act upon that environment through actuators, processing information and knowledge. A multiagent system (MAS), in turn, consists of a set of autonomous agents that collaborate to solve a problem which would be impossible to solve with just one agent.
Agents can be constructed in various ways. They can be agents of software or hardware, static or mobile, per- sistent or non-persistent, reactive or cognitive (intelli- gent). According to [1], one of the most important classifi- cation of agents is in relation of them to be reactive or cognitive. Reactive agents are agents that select actions to perform based solely on current perception, not consider- ing the historical of perceptions. Since they do not have memory, they are unable to plan future actions. The cog- nitive agents are more complex since they have an explicit representation of both the environment and the other agents. This agent type has memory, which enables it to plan future actions based on situations that took place previously [8], [1].
Also according to [8], another feature to be taken into consideration is the rationality, which is influenced by four factors: i) performance measure (it defines the suc- cess criteria); ii) prior knowledge of the agent; iii) actions that the agent can perform; and iv) following perceptions captured by the agent so far. Faced with these elements, it is possible to conceptualize a rational agent as one which, for each possible sequence of perceptions, selects an ac- tion that will maximize its performance measure, given the evidence provided by the sequence of perceptions and any internal knowledge of the agent [8]. Considering this definition, one can realize that is not always that a ration- al agent will make the best decision possible, but one that improves its performance measure.
Intelligent agents can perform various tasks in context- aware learning environments, such as i) monitoring the activities of the student in the learning environment, ii) automatically capture information from the dynamic context of the student, iii) recommend interest content for that student, among others. Before the increase in the number of students who interact with learning support systems, the use of agents to perform these tasks become extremely important, mainly due to the fact that they are complex tasks for the facilitators to manage from dis- tance.


An important concept regarding the educational content used in DE is the Learning Object (LO). According to Learning Technology Standard Committee - LTSC, from the Institute of Electrical and Electronics Engineers (IEEE), a LO is an educational material entity, digital or not, that can be used for learning, education or training [7].
According to [9], there are seven characteristics of LOs, while for [10] there are four properties presented by LOs.
Table 1 shows a comparison between the LOs properties presented by these authors.



Thus, a LO, beyond the content itself, has a structure containing metadata that allows, through its elements and attributes, describing its contents, its format, how that content is presented, beyond other information (content author, creation date, educational data, etc.). This metada- ta is created based on standards that will be presented in Subsection 3.1.
In summary, the central idea of the concept of LOs is to allow educational designers build relatively small educa- tional components that can be used in different learning contexts. In other words, they are digital contents that enable or facilitate reaching an educational goal and their reusability [11].

3.1 Learning Objects Standards

Despite the benefits of using LOs, it should be also consi- dered the problems faced in creating digital LOs. [5] de- scribes a series of difficulties that are faced while creating digital LOs: i) definition of the navigational structure; ii) adequacy of the contents of a print media to electronic media; iii) assistance to the pedagogical aspects of teach- ing; iv) integration of LO with different types of DE envi- ronments; and v) high cost of authoring tools’ licenses.
To solve these problems, LOs standards are used. Ac- cording to [12], these standards are a way of organizing the data of a LO to provide communication among differ- ent computing environments, as well as ensure its acces- sibility and usability, and also to provide interoperability. The authors report that these patterns are divided accord- ing to their functionality in: metadata, packaging, inter- face and communication and integration standards.
The metadata standards are used in the resources’ identification, aiding in the filtering of a search and re- trieval of a record or a LO [12]. An example of the meta-

IJSER © 2011

International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 3

ISSN 2229-5518

data’s standard is the LOM (Learning Object Metadata) [7], which was developed by LTSC [6]. The purpose of this standard is to facilitate search, evaluation, acquisition and use of LOs, both by students and instructors, or even automated software processes. Additionally, this stan- dard also facilitates the distribution and exchange of LOs, allowing the development of catalogs and inventories considering the diversity of linguistic and cultural con- texts in which the LOs and their metadata are reused. By specifying a common data schema, this standard ensures that the connections of LOs have a high degree of seman- tic interoperability [7]. Data elements describe a LO and they are grouped into categories. The basic metadata’s structure of the LOM is defined in nine categories, as shown in Table 2 [7].



An integration standard, as the name implies, unifies in a reference model different types of standards, such as metadata standards, packaging, communication and in- terface, developed by other organizations [12]. The inte- gration standard SCORM (Sharable Content Object Refer- ence Model), developed by ADL (Advanced Distributed Learning) [10], integrates a set of technical standards, specifications and guidelines designated to attend the requirements of high level of SCORM - systems and ac- cessible content, interoperable, durable and reusable. The content in SCORM standard can be distributed to stu- dents through any LMS that is compatible with and use the same version of SCORM [10].
In this paper we will use LOs developed according to
the SCORM standard, both because it is a widespread standard and it is also composed of a series of other stan- dards. Thus, we can enjoy the best benefits offered by each standard.


An ubiquitous learning environment may be understood as a context-aware mobile learning environment, provid- ing most adaptive contents for learners. Context aware- ness describes a paradigm in which the context of a user is considered to define his profile [3], [14]. There is no consensus about the definition of ―context". This one is specific of the application and the desired intention, re- quiring the identification of functions and properties of the individuals’ domains [3], [13], [14].
Context can be defined as information consisting of
properties that combine each other to describe and cha- racterize an entity and its role as a computer-readable
form [3], [14].
The location is crucial to the context of the student in
an environment for ubiquitous learning. However, the
context includes more than just the location. A wide
range of context factors combine themselves to form a context definition. Almost all information available at the
moment of interaction can be seen as contextual informa- tion, among which stand out [3], [15]:
• The various tasks required from users;
• The wide range of devices that combine to create
mobile systems with associated infrastructure services;
• Resources availability (e.g. battery status, screen size,
network bandwidth, etc.);
• Resources in the neighborhood (e.g. accessible devic-
es and servers);
• The physical situation (e.g. temperature, air quality,
brightness level, noise etc.);
• Spatial information (e.g. location, orientation, veloci-
ty, acceleration etc.);
• Time information (e.g. time of day, date, season,
• Physiological measures (e.g. blood pressure, heart
rate, respiratory rate, muscle activity etc.).
The list above, though not exactly contain all the in-
formation that can be considered, is used to demonstrate
the inherent complexity of the context, its specific nature
of domain and difficulty to define and measure it [3]. In
an attempt to reduce this complexity, [14] defines two
general types of context: i) static context (named customi- zation), which concerns a use case in which a user profile
(context) is created manually and the user is actively in- volved in the process and having an element of control; and ii) dynamic context (named personalization), which refers to the condition in which the user is seen as pas- sive, or at least with a little less control. In this case, the system monitors, analyzes and reacts dynamically to user behaviour and the identified role.
Many context-aware mobile learning applications

IJSER © 2011

International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 4

ISSN 2229-5518

(ubiquitous learning environments) use learning contexts in order to appropriately adjust or suggest content and activities for students [2]. However, these works do not consider the capabilities of mobile devices, what com- promises a more precise definition of the context of stu- dents.


Multiagent systems (MAS) have been widely used in educational environments. This technology can provide an aid to learning environments, making these environ- ments more proactive and autonomous. MAS can be use- ful, for example, in developing a context-aware feature in a learning environment.
In [2], it is described a context-aware mobile learning environment architecture consisting of three main com- ponents: the learner profile, a personalization mechanism and a LO repository. The learner profile stores his mobile learning preferences, gotten through a questionnaire answered by the learner, in which he indicates preferred location of study, noise level and time of day. The perso- nalization mechanism gets this profile and matches its information with the current values detected by the con- text-aware technologies. Then, the personalization me- chanism compares all information obtained with the me- tadata of LOs available in the repository. The system then recommends appropriate LOs for students in accordance with their context characteristics.
In [16], it is described an ubiquitous learning architec-
ture for supporting student to learn English as foreign
language in order to prepare for TOEFL (Test of English
as a Foreign Language). The system provides adaptive
content for different learners based on context-awareness,
considering location, time, manner as well as learner’s
knowledge. This is possible through suggested topics as well as test questions. [16] also describes CAMLES (Con- text-Aware Mobile Learning English System) prototype, that allows the learner to receive adaptive materials for TOEFL anytime in anywhere with mobile phone. The main difference of our work for the last one is the fact that our approach considers the definition of context, beyond the student’s profile information, information of available physical resources such as connection type, format (video, audio, etc.) supported by the mobile devices.


The agent-based approach proposed in this paper is pre- sented in Figure 1. According to this approach, three types of agents are proposed: Student Agent (SAg), Re- commender Agent (RAg) and Interface Agent (IAg). The SAgs are responsible for monitoring the activity of stu- dents in the learning environment and sending to RAgs static and dynamic information of the student’s profile.
These information are recovered from the ontology of static and dynamic context respectively. The RAgs are responsible for identifying appropriate learning objects (LOs) to the student’s context, according to both the in- formation provided by SAgs and the learning objects available in the repository. Once identified these informa- tion, the RAgs inform both the IAg and the instructor.

Fig. 1. Agent-based approach for recommending learning objects

6.1 Agents Model

This section describes how the agents in the proposed agent-approach were modeled and implemented. For a better understanding, it was chosen to detail this ap- proach using the MAS - CommonKADS+ [18], a metho- dology for modeling multiagent system, which consists in an extension to the traditional methodology MAS- CommonKADS.

This methodology describes the agents through dia- grams and CRC (Class-Responsibility-Collaborators) cards, referring to the tasks of the system. These artifacts describe features that the system should perform. The roles model shows the roles which each agent can play in the system. These roles are responsible for meeting the existing tasks in the task model. The diagram in Figure 2 shows the system roles model.

Fig. 2. Roles model

CRC cards, on the other hand, show the details of the agents and their goals. Each CRC card is built to specify a particular agent. The agent goals are listed and the plans

IJSER © 2011

International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 5

ISSN 2229-5518

(actions that should be performed to achieve the goals) are described. It also identifies the knowledge that the agents own to perform their actions and the agents that collaborate in the execution of plans and services through which the interaction takes place among agents [19]. The CRC cards will be shown in the next sections.
Another model used in this methodology is the agent model. This one is responsible for defining which agents will be responsible for each presented role, by defining the agent architecture, their goals and characteristics, such as data entry, activation conditions of the agent and available information types. Figure 3 shows this diagram for the Recommender Agent (see Section 6.3).

Fig. 3. Agent Model

The MAS-CommonKADS+ methodology, in its turn, proposes the integration of AML (Agent Modeling Lan- guage) to the MAS-CommonKADS methodology. Finally, it also suggests, among other things, the use of an interac- tion model to describe all the interactions among agents through the AML. This model is represented by interac- tion diagrams, like sequence and communication dia- grams, using AML diagrams [18]. It was chosen to use the sequence diagram in this work. This diagram will be shown in the next sections.

6.2 Student Agent (SAg)

SAgs are intended to search, in both the static and the dynamic context ontologies of the students, the learning preferences that make up their profiles and dynamic in- formation of the student context. To facilitate the com- prehension of the behaviour of SAg, Table 3 shows the CRC card of the SAg.
finding any student, the agent remains in a locked state. If a new student has logged into the LMS, the agent loads the student profile ontology model, and it looks for the preferences of this student. Then, the preferences infor- mation, in conjunction with information about the stu- dent geographical location, are sent to the RAg, so that it can identify LOs tailored to the needs of this student.
The following Java code shows the SAg’s behaviour whose aim is to read both the static and the dynamic student information.

1. public class StudentAgent extends Agent{

2. [...]

3. private class SearchForStudent extends OneShotBehaviour{

4. @Override

5. public void action(){

6. int id = (int) getArguments()[0];

7. Location location = (Location) getArguments()[1];

8. Student student = new OntologyParser(ONTOLOGY_PATH).

9. getStudent(id);

10. student.setLocation(location);

11. send(student);

12. } // End of action() method

13. } // End of SearchForStudent Behaviour

14. [...]

15. } // End of StudentAgent

In the line 6 of the presented code, the SAg discovers the current location of the student. In the line 8, it maps information from the student’s preferences in the ontolo- gy through a Java class, according to the specified student ID. Finally, in the line 11, it sends the student profile to RAg. Figure 4 shows the messages exchange flow be- tween the SAg and the RAg through the interaction mod- el.



The SAg has a one shot behaviour, in charge of perce- ive the authentication of a student at LMS. While not

Fig. 4. Interaction model between the SAg and the RAg.

6.3 Recommender Agent (RAg)

The RAg is responsible for detecting LOs appropriated to the context in which the student is inserted, according to information provided by the SAgs and the metadata in- formation of the available LOs in the repository. Once the student context characteristics have been identified, it communicates to the instructor and passes information

IJSER © 2011

International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 6

ISSN 2229-5518

from the LO to the Interface Agent (IAg). To facilitate the



comprehension of the behaviour of the RAg, Table 4


new MessageDAO().sendMessage(2, "Current location:

shows the CRC card of the RAg. The personalization


City: "+student.getLocation().getCity()+"

mechanism used by the RAg is presented in Subsection



The following code shows the Java code of the RAg cyclic behaviour, which goal is to wait the student pro- files sent by SAg.

1. private class WaitStudent extends CyclicBehaviour{

2. @Override

3. public void action() {

4. try{

5. ACLMessage msg = myAgent.receive();

6. if(msg!=null){

7. student = (Student) msg.getContentObject();

8. addBehaviour(new FindLearningObjects());

9. }else{

10. block();

11. }

12. }catch(Exception e){

13. System.out.println("Exception in


14. " +e.getMessage());

15. }

16. } // End of action() method

17. } // End of FindLearningObjects Behaviour

In the line 6 of the presented code, the RAg verifies if there is any message in its messages’ queue. If there is any message, the RAg makes an instance of the student according to this content (line 7). Following, in the line 8, it executes the FindLearningObjetcs behaviour to find the suitable LOs. If there is no message, the RAg keeps itself blocked until a new message shows up (line 10).
The next code shows the FindLearningObjects beha-
viour, which is a kind of one shot behaviour. It is respon-
sible for finding and recommending suitable LOs, accord-
ing to both the student profile and the dynamic context
information of the student.

18. private class FindLearningObjects extends OneShotBehaviour{

19. @Override

20. public void action() {

21. for(LearningObject recommendedOA:new RBC(student).

25. State: "+student.getLocation().getState()+"

26. Country: "+student.getLocation().getCountry()+"

27. <br/>LearningObject:”

28. "+recommendedOA.getTitleTagValue()[0]);

29. } // End of for loop

30. } // End of action() method

31. } // End of FindLearningObjects Behaviour

How can be viewed in the presented code, for each suitable LO found, the RAg sends a message to the stu- dent through the mobile application interface (lines 21 -

29). Figure 5 shows the messages exchange flow between the RAg and the IAg through the interaction model.

Fig. 5 Interaction model between the RAg and the IAg.

The Figure 6 shows the recommendation performed by this behaviour. The Subsection 6.3.1 shows how the per- sonalization mechanism is performed.

Fig. 6. Main mobile application interface to recommend LO.

6.3.1 Personalization Mechanism

The personalization mechanism considers the pre- ferred location and time for studying and preferred inter- est area of student, which can be found in the static con- text ontology. These information are weighted according to the influence level that each one implies in the student learning model. The strategy to identify the suitable LOs

IJSER © 2011

International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 7

ISSN 2229-5518

according to the student context is performed in accor- dance with the Equation 1.


The Recommendation Factor (RF), which is deter- mined by the RAg, is mainly influenced by the Interest Area (IA) of the student, having therefore weight 5. The Preferred Location (PL) of study receives weight 3, since it has enough influence in the student’s concentration capability. Finally, the Preferred Time (PT) of study is also of interesting, because it has influence in the concen- tration level and, thereafter, it can influence in a positive or negative way on the LO recommendation. The Ad- justment Factor (AF) regards to a factor that can be estab- lished by the instructor in order to increase (when AF is smaller) or decrease (when AF is greater) the impact that the user’s context implies in the LO recommendation. The numeric values of IA, PL and PT represent a comparison between the dynamically found values and the ones which have been defined in the student static context ontology.
To dynamically define the value which represents how suitable is the LO regarding to interest area of a student, the LO characteristics have three considerations: descrip-
tion, title and keywords. The RAg, on the other hand,
verifies the frequency of student’s interest words, which
are present in the static context ontology, in these three LO characteristics. On following, these values are weighted by the RAg according to the Equation 2.


The Equation 2 shows that the Keywords (KW) have the greatest weight, whereas they represent the issues treated in the LO in a better way. The Description (D) of the LO gives us an overview of how its several issues are integrated. Finally, the Title (T) represents a smaller influ- ence among the three characteristics, due to the fact it does not contain a range of related words as extensive as the KW.
To dynamically define the PL factor, it has been made an extension in the application development framework MLE (Mobile Learning Engine) [20]. It has been added to this framework a class which is in charge of finding the student geographical location from an API (Application Programming Interface) integrated to the GPS (Global Positioning System) of the mobile device. Thus, it is poss- ible to acquire the latitude and longitude of the mobile device. Then, this information is sent by the application client of the mobile device to SAg. On the other hand, the SAg performs the reverse geocoding3 of information before sending them to RAg.
Finally, the SAg dynamically finds the current time at the moment of system student authentication. This in- formation will be used by the RAg to define the numeri- cal value which will be assigned to the PT.

6.4 Interface Agent (IAg)

The Interface Agent (IAg) aims to provide students with information as they progress in the distance learning course in which they are participating. This agent adjusts the type of message according to the device capabilities of the student. To facilitate the comprehension of the IAg’s behaviour, Table 5 shows the CRC card of the IAg.




In this paper, it was described the implementation of an agent-based approach for recommending learning objects in ubiquitous learning environments. The pro- posed solution aims to make the learning environment suitable to the student’s needs, considering their contexts. The presented agent-based approach can be used together with any learning management system, once it has been developed as an software layer independent of the appli- cation.
As further work, we can consider other user informa-
tion, like the courses which the student has already parti-
cipated in the past. This way, system could suggest LOs
according to these preferences.
Another issue that can be implemented is the collabor-
ative recommendation. This kind of recommendation
considers suitable LOs according to similar preferences of
students. First of all, the system identifies students with
similar preferences. Then, it considers LOs chosen by the
students in the past. According to these choices, the sys- tem suggests suitable LOs to the present student.
We also intend to do a case study with undergraduate students of the computer science course to evaluate the impact of the proposed agent-based approach in the learning content suitability.


[1] A.A.A. Pontes, F.M. Mendes Neto and G.A. Campos, ―Multiagent System for Detecting Passive Students in Problem-Based Learning,‖ Trends in Practical Applications of Agents and Multiagent Systems, vol. 71, pp. 165–172, Heidelberg, Springer Berlin, 2010.

[2] J.Y.K. Yau and M. Joy, ―A Context-Aware Personalized M- learning Application Based on M-learning Preferences‖, VI IEEE International Conference on Wireless, Mobile and Ubiquitous

IJSER © 2011

International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 8

ISSN 2229-5518

Technologies in Education, (VI WMUTE), pp. 11-18, Washington, USA, IEEE Computer Society, 2010.

[3] P. Moore et al, ―Intelligent Context for Personalised M-

Learning‖, International Conference on Complex, Intelligent and Software Intensive Systems, pp. 247–254, Los Alamitos, USA, IEEE Computer Society, 2009.

[4] X. Zhao, X. Whan and T. Okamoto, "Adaptive Content Delivery in Ubiquitous Learning Environment", VI IEEE International Conference on Wireless, Mobile and Ubiquitous Technologies in Edu- cation, (VI WMUTE), pp. 19-26, Washington, USA, IEEE Com- puter Society, 2010.

[5] E.R. Rodolpho, "Convergência digital de objetos de

aprendizagem SCORM", Master Thesis, Instituto de Biociências, Letras e Ciências Exatas, Universidade Estadual de São Paulo Júlio Mesquita Filho, São José do Rio Preto, Brazil, 2009.

[6] Learning Technology Standards Committee, LTSC, available at:

<>, acessed in: 15 abr. 2011.

[7] "Draft Standard for Learning Object Metadata IEEE Standard

1484.12.1", available at:


raft.pdf>, accessed in: 18 abr. 2011.

[8] S.J. Russel and P. Norvig, Artificial Intelligence, a Modern Ap- proach, 2nd Edition. New Jersey, USA: Prentice Hall, 2003.

[9] S.E. Caregnato, R.M. Mendes and V.I. Souza, "A propriedade intelectial na elaboração de objetos de aprendizagem", V Encontro Nacional de Ciência da Informação, Salvador, Brazil, Instituto de Ciência da Informação, 2004.

[10] Advanced Distributed Learning, ADL, available at:

<>, acessed in: 20 abr. 2011.

[11] S. Castillo and G. Ayala, "ARMOLEO: An Architecture for Mobile Learning Objects", 18th International Conference on Elec- tronics, Communications and Computers (18th CONIELECOMP), pp. 53-58, Los Alamitos, USA, IEEE Computer Society, 2008.

[12] C.C.L. Dias et al, ―Padrões abertos: aplicabilidade em Objetos

de Aprendizagem (OAs)‖, XX Simpósio Brasileiro de Informática na Educação (XX SBIE), Florianópolis, Brazil, 2009.

[13] P. Moore and B. Hu, "A Context Framework with Ontology for Personalised and Cooperative Mobile Learning", International Conference on Computer Supported Cooperative Work in Design III, pp. 727-738, Berlim, Springer-Verlag, 2007.

[14] P. Moore, B. Hu and J. Wan, ―Smart-Context: A Context Ontol- ogy for Pervasive Mobile Computing‖, The Computer Journal, vol. 53, pp. 191-207, 2008.

[15] B. Schilit, N. Adams and R. Want, "Context-Aware Computing Application", IEEE Workshop on Mobile Computing Systems and Applications (WMCSA), pp.85-90, Santa Cruz, USA, IEEE Com- puter Society, 1994.

[16] V.A. Nguyen, V.C. Pham and S.D. Ho, "A Context-Aware Mobile Learning Adaptive System for Supporting Foreigner Learning English", International Conference on Computing and Communication Technologies, Research, Innovation, and Vision for The Future (RIVF), pp. 1-6, Hanoi, Vietnam, IEEE Computer So- ciety, 2010.

[17] C. Yao, "Adaptive context aware and intelligent searching in

mobile learning applications", 2nd IEEE International Conference on Computer and Automation Engineering (ICCAE), pp. 802-806, Taipei, Taiwan, IEEE Computer Society, 2010.

[18] M.J.O. Morais II, "MAS-Commonkads+: Uma Extensão à Metodologia MAS-Commonkads para Suporte ao Projeto Detalhado de Sistemas Multiagentes Racionais", Master Thesis, Universidade Estadual do Ceará, Fortaleza, Brazil, 2010.

[19] F.M. Mendes Neto and M.J.O. Morais II, "Multiagent System for Supporting the Knowledge Management in the Software Process", Muthu Ramachadran (Org.), Knowledge Engineering for Software Development Life Cycles: Support Technologies and Appli- cations, pp. 96-113, New York, IGI Global, 2010.

[20] Mobile Learning Engine, MLE, available at:

<>, acessed in: 30 abr. 2011.

IJSER © 2011