Adaptive Learning Objects for T-learning Marta Rey-López∗ Ana Fernández-Vilas∗ Rebeca P. Díaz-Redondo∗ José J. Pazos-Arias∗ Jesús Bermejo-Muñoz† ∗
Department of Telematic Engineering, University of Vigo, Spain {mrey,avilas,rebeca,jose}@det.uvigo.es †
TELVENT Interactiva S.A., Sevilla, Spain
[email protected]
Abstract IDTV(Interactive Digital TV) opens new learning opportunities for those social groups that would hardly have access to traditional forms of education. However, viewers are not usually active learners, for this reason, education through IDTV should be offered in an attractive way, so as they get engaged in the learning experience. For this to be possible, we need to introduce personalization in the t-learning field. To achieve this goal, we present, in this paper, a proposal of self-adaptive t-learning objects, which show a different behaviour depending on user’s characteristics. These objects are conformant to the ADL SCORM (Sharable Content Object Reference Model) standard for which we propose an extension in order to permit the learning objects to adopt different behaviours. We expose as well an authoring tool to create this kind of learning objects which hides the content creator the implementation details. Keywords: SCORM, t-learning, adaptive SCOs.
1. Introduction The growing success of IDTV (Interactive Digital TV) offers new services that have not traditionally been associated with this medium, such as commerce or learning. The term t-learning has been adopted [2] to denote TV-based interactive learning, which is not just an adaptation for IDTV of the e-learning techniques used in the Internet. It has its own distinctive characteristics, mostly related to the constraints imposed by the television set and the set-top box, such as the distance between the student and the screen —which makes more difficult reading a text— or the fact of using a simple remote control to operate them —which reduces the possibilities of interaction with the student. Furthermore, applications are executed on a set-top box which has lower computer power than a personal computer, that is why application’s complexity is inexorably reduced. These characteristics impose a different conception of learning objects. As opposed to e-learning ones, they should principally consist of audio and video (traditionally used in television) reducing the appearance of text to the minimum, since it will be difficult for the user to read it. To develop t-learning objects we should also take into account social characteristics. The most relevant one is the predisposition of the student towards education. In e-learning, the student gets generally involved in learning experiences on his own initiative. On the contrary, the t-learning student, who has traditionally been a viewer and used TV solely as an entertainment medium, is usually more passive. For this reason, personalization is essential in IDTV, in order to make t-learning objects more attractive and effective for the
user. More attractive because they adapt to the learner’s preferences, and more effective since they take into account his/her learning background, as well as the goals he/she wants to achieve. With the aim of solving the problem of personalization in t-learning, we present self-adaptive t-learning objects. Their main feature is being able to change their behaviour according to the concrete characteristics of the student. The main component of these objects, in order to show the most appropriate behaviour, is an adaptation file. This file brings t-learning objects the adaptation rules that they need to know which appearance has to be shown to the target user, according to his/her preferences and learning background. These characteristics have been restricted to a set of adaptation parameters, which consist of the most relevant user’s characteristics in a concrete domain. This proposal takes place into a broader work whose objective is developing an Intelligent Tutoring System (ITS) for t-learning, called t-MAESTRO (Multimedia Adaptive Educational SysTem based on Reassembling TV Objects) [8]. Its principal novelties with respect of the ITSs for e-learning are that it works for a user with special characteristics because he/she is a student and a viewer at the same time and it composes media driven courses, based on video and audio content. This ITS is designed to work over a Multimedia Home Platform (MHP) [5] with learning material conformant with the ADL SCORM (Sharable Content Object Reference Model) [1] standard. In SCORM terminology, what we have been referring to as learning object is called SCO (Sharable Content Object), which represents a single launchable object that can communicate with the LMS (Learning Management System)1 . Bearing this in mind, the self-adaptive t-learning objects aforementioned are hereinafter denominated self-adaptive SCOs. In this paper, we expose a solution to offer personalization in the t-learning field: self-adaptive SCOs. The scenario where these learning objects should work is shown in Sec. 2. Next, we explain the importance of adaptation parameters (Sec. 3) and expose the structure and operation of the self-adaptive SCOs (Sec. 4). To make the creation of these objects easier, we present SCOCreator, an authoring tool to prevent the creator from knowing about programming (Sec. 5). Last but not least, we discuss the related work with respect to our proposal (Sec. 6) and expose the conclusions and future lines of our work in Sec. 7.
2. Scenario of our Proposal As mentioned above, the self-adaptive t-learning objects we are presenting are SCORM-conformant, that is why we call them —in accordance with the SCORM terminology— self-adaptive SCOs. SCORM references specifications, standards and guidelines developed by other organizations that are adapted and integrated with one another to form a more complete and easier-to-implement model. It is divided into technical books grouped under three main topics: SCORM Content Aggregation Model (CAM), covering assembling, labelling and packaging of learning content; SCORM Sequencing and Navigation (SN), describing how educational content may be sequenced through a set of navigation events; and SCORM Run-time Environment (RTE), whose purpose is providing a means for interoperability between SCOs (Sharable Content Objects) and LMSs. In this paper, we are specially interested in SCORM RTE since it permits the SCO to exchange information with the LMS and change its behaviour accordingly. To establish this communication, SCORM RTE defines an API, whose most relevant method for this proposal is GetValue(). This method is used by the SCO to request information stored in the SCORM Data Model, which is also defined in the SCORM RTE book. This model stores information mainly related to the completion of the objectives of the activities that compose the course and does not keep external information. 1
In e-learning terminology, the term LMS is used to refer to the system designed to deliver, track, report on and manage learning content, learner progress and learner interactions.
op2
op3
Self-adaptive op1
SCO
op4
User
Content creator
LMS
t-MAESTRO
User profile
Inference rules
Vocabulary of adaptation parameters
SCORM data model
Extended Data Model Vocabulary creator
Scenario defined by SCORM
Figure 1. Scenario for self-adaptive SCOs In this manner, the standard does not permit the LMS to communicate the user’s characteristics to the SCO, since they are external to the current course. For this reason we propose an extension of the SCORM standard, concretely of the SCORM Data Model with some relevant user’s characteristics that we have called adaptation parameters (adaptation parameters will be further explained in Sec. 3). We can see in Fig. 1 the complete scenario where the self-adaptive SCOs will be launched, in it, the SCORM Data Model has been extended with the vocabulary of adaptation parameters. In this figure, we can see three roles in the adaptation process: the vocabulary creator, the content creator and the user; as well as the two systems needed to communicate the user’s characteristics to the SCO: the LMS and t-MAESTRO. First, the vocabulary creator extends the SCORM Data Model with new vocabularies of adaptation parameters and provides t-MAESTRO with inference rules. These rules permit to obtain the actual values of these parameters from the user’s characteristics stored in the user profile. They also make possible that this proposal works independently of how the user profile is stored in the ITS. Next, the content creator creates SCORM-conformant self-adaptive SCOs (using, for example, the SCOCreator tool presented in Sec. 5) with their different behaviours, as well as the adaptation rules it needs to decide which behaviour it has to adopt according to the user’s characteristics. t-MAESTRO maintains the user profile, which stores the user’s preferences, knowledge and history. The existence of this profile is basic in order for the ITS to keep up to date the actual values of the adaptation parameters using the inference rules for these parameters provided by the vocabulary creator. Finally, the LMS has the responsibility of storing the actual values of the adaptation parameters and showing the SCO to the user. This system has to provide the values of adaptation parameters to self-adaptive SCOs when requested —by means of the GetValue() method—, so as they can resolve the adaptation rules and show the appropriate behaviour to the student.
Adaptation parameters t-maestro.preferredResourceType DT (Data type): state (audio, image, text, video) t-maestro.preferredSport DT: state (basketball, f1, football, etc.) t-maestro.preferredTVProgramType DT: state (documentaries, magazines, etc.) t-maestro.preferredMoviesGenre DT: state (adventure, animation, comedy, etc.) t-maestro.preferredMusicGenre DT: state (orchestral, pop, ragtime, rap, etc.)
t-maestro.typeOfDisability DT: state (none, hearing_impaired, etc.) t-maestro.languages (user’s level in different languages) t-maestro.languages.n.language DT: langcode t-maestro.languages.n.listening_level DT: Non-negative integer t-maestro.languages.n.speaking_level DT: Non-negative integer t-maestro.languages.n.reading_level DT: Non-negative integer
t-maestro.age
t-maestro.languages.n.writing_level
DT: Non-negative integer t-maestro.academic_orientation DT: state (science, technology, culture, etc.)
DT: Non-negative integer t-maestro.subjects (user’s level in different subjects) t-maestro.subjects.n.subject
t-maestro.mother_tongue
DT: state (chemistry, history, maths, etc.)
DT: langcode t-maestro.educationalLevel DT: state (primary_school, high_school, etc.)
t-maestro.subjects.n.level DT: Non-negative integer
Table 1. Proposed adaptation parameters
3. Adaptation Parameters As we have introduced in previous sections, adaptation parameters consist in some relevant user’s characteristics. In this section, we present a set of adaptation parameters which are appropriate for t-learning and we explain how t-MAESTRO (and other possible ITSs) can deduce the actual values for these parameters from the information stored in the user profile. In Table 1, we can see some adaptation parameters that we consider appropriate for a t-learning student, such as the movies genre he/she prefers or his/her level in different subjects. The actual values for these parameters could be directly stored in the user profile or should be inferred from some other user’s data in this profile. Let us expose an example: the ITS wants to know the user’s level in maths. It can have this value stored in the user profile —so it already knows it— or it can deduce it from other information, for example, if the learner is an engineer, the ITS can infer that his/her maths level is high. In order to help the ITS to establish the relationships mentioned above between the information in his/her user profile and the values for the adaptation parameters, the vocabulary creator should provide it with inference rules. These inference rules are different for each ITS and allow the content creator to create self-adaptive SCOs independently of the user profile stored in the ITS.
4. Self-Adaptive SCOs We have already mentioned in Sec. 2 that SCORM RTE permits an SCO to exchange information with the LMS as long as this information is stored in the SCORM Data Model. To produce self-adaptive SCOs, we take advantage of this exchange of information and propose an extension to this data model with a vocabulary of adaptation parameters containing some relevant user’s characteristics (Sec. 3).
4.1 Structure The SCORM standard does not restrict the internal operation of a SCO on condition that it communicates with the LMS using SCORM RTE. Taking this fact into account, we propose the structure in Fig. 2 for selfadaptive SCOs, where we can observe their three main components: a Java template, several configuration files and an adaptation file. SCO
.
JAVA template
.
Configuration file n
Figure 2. Structure of a self-adaptive SCO
Destroyed
initXlet()
ro st de
t le yX
()
destroyXlet()
Paused pauseXlet()
.
destroyXlet()
Adaptation file
Configuration file 2
Loaded
startXlet()
Configuration file 1
Active
Figure 3. Life cycle of an Xlet
The Java template contains the SCO functionality, for example, a Java class with some space for text, a video, a picture and some control buttons. The actual objects taking up these spaces are loaded on runtime. In order for this Java class to work over a MHP context, it should be conformant with the DVB-J model defined in the standard. DVB-J applications, commonly referred to as Xlets, are Java programs with two restrictions: the APIs they are allowed to use and its life cycle. The Xlets should follow a life cycle that allows an external entity to initialize, pause, restart and destroy them by means of an established interface, as shown in Fig. 3. Each configuration file is an XML file which specifies the behaviour of the SCO for a concrete option. It indicates which objects take up the spaces in the template as well as the properties of these objects, such as colour, position, etc. There are as much configuration files as the possible different behaviours the SCO can adopt. The syntax of these files has no restrictions as long as the SCO understands them. In our case, we have defined a syntax where all the properties of the objects of the template that have been modified are listed. Last but not least, the adaptation file is an XML file which contains the adaptation rules that indicate the SCO which is the most appropriate behaviour to adopt. These rules can be resolved according to the actual values of the adaptation parameters, in order to know which is the configuration file it has to use for a concrete execution. The syntax of the adaptation files, although it has been already defined, is out of the scope of this paper. 4.2 Operation We can see in Fig. 4 the operation of a self-adaptive SCO when launched by the LMS. First, it reads the Adaptation File with the purpose of knowing which are the Adaptation Parameters it needs to know so as to resolve the Adaptation Rules (step Ê). Next, it requests the LMS the actual values of these parameters for the concrete user, using consecutive calls to the GetValue() method of the SCORM RTE API (step Ë). With these values, it resolves the Adaptation Rules in the Adaptation File in order to know which Configuration File to use (step Ì). The Configuration File contains the properties of all the objects in the Java Template of the SCO, so as it can take the appropriate behaviour for the target user (step Í). With this behaviour, the SCO interacts with the user to carry out its educational mission (step Î). In the example shown in the figure, the SCO belongs to a Spanish course. Its mission is helping the learner with listening comprehension. A video or audio —depending on his/her preferences— with subtitles
n
o
Java template
Java template GetValue(“preferredResourceType”)
Adaptation file
preferredSport preferredResourceType
LMS
video GetValue(“preferredSport”) tennis
Subtitles
p
q
Java template Adaptation file
preferredSport: tennis preferredResourceType: video
Configuration file: tennis_v.xml
Subtitles
Java template
Configuration file
Subtitles
Subtitles
r
User
Figure 4. Operation of a self-adaptive SCO is shown to the student. The content of this multimedia object depends on the user’s favourite sport, with the intention of being interesting for him/her. The Adaptation Parameters that the SCO needs to resolve the Adaptation Rules are preferredResourceType (audio or video) and preferredSport (tennis, football, basketball. . . ). When it requests these values to the LMS, this one returns video and tennis respectively. It resolves the Adaptation Rules and, according to the Configuration File, it shows a video with a portion of a tennis match with its subtitles. These are some of the rules for this example in pseudocode: if ((t-maestro.preferredSport == football) AND (t-maestro.preferredResourceType == audio)) then //Rule action file = football_a.xml; if ((t-maestro.preferredSport == tennis) AND (t-maestro.preferredResourceType == video)) then //Rule action file = tennis_v.xml;
5. The SCOCreator Tool In this section we present SCOCreator, an authoring tool that we have developed to create self-adaptive SCOs from Java templates. With this tool, the content creator should be able to design the appearance of the SCO for each of its possible different behaviours (hereinafter called options), as well as the configuration and adaptation files for this SCO, without knowing anything about Java programming or understanding and writing XML files. The first step in the process of creating the self-adaptive SCO is choosing the appropriate template from those ones that have been previously created. As we have already mentioned in Sec. 4.1, these templates
Figure 5. Option 1 consist in Java classes with the Xlets restrictions, which contain several multimedia objects, such as video, audio, text or images. Once the template is loaded, we can distinguish three different areas on the screen (see Fig. 5). On the left, the template is shown, where all its objects can be mouse selected in order to change their properties. On the top-right corner, the content creator can change the values of the properties of the selected object for each of the options he/she wants to provide the SCO with (a tab is offered for each option). These tabs do not permit changing the size, position, depth and visibility of the object, which should be changed using the toolbar above the template. Finally, on the bottom-right corner, we can see the objects hierarchy which makes easier to select objects that are contained in others. To produce the self-adaptive SCO, the content creator should create the options he/she wants to provide the SCO with and change its objects’ properties for every option, in accordance with the characteristics of the target user the option is designed for. For example, if the template is composed by a media player (with the buttons it needs) and a label to show the subtitles for the audio or video file offered in the player, one of the options can be designed for people with eyesight difficulties, so, the text font should be changed into a larger one (Fig 6). All the properties that have been changed for a concrete option are stored in the configuration file for this option, which consist in a XML file that the self-adaptive SCO reads to adopt the appropriate behaviour for this option. After designing each option, the content creator should define the adaptation rules, in order to indicate the SCO which behaviour it has to show to each user according to his/her characteristics, i.e. which configuration file it has to use in the configuration phase. In order to make this task easier, an adaptation rules editor has
Figure 6. Option 2 been designed (Fig. 7). On it, the content creator should be able to indicate the characteristics of the target user for each option using a logical expression. In the example given above, option 1 should be offered to a user with no disability, while option 2 is intended for users with visual impairments. When clicking on the “Validate” button, the program converts the logical expression on XML code, so as the creator does not need to know the syntax of the adaptation files. The last step is creating the self adaptive SCO, in this moment, the SCOCreator tool generates the configuration file for each option, the adaptation file and gathers the Java classes and other files —e.g. video, images or audio files— needed for the SCO to work properly.
6. Related Work One of the most promising fields to offer personalization on the e-learning field is Adaptive Hypermedia (AH), which pursues the adaptation of hypermedia documents, i.e. those where different media are used and where the user has many ways to navigate between different information objects. As Peter Brusilovsky states in [4], AH tries to overcome the problem of having users with different goals and knowledge by using information represented in the user model to adapt the contents and links being presented to the given user. AH systems can be useful in any application area where the system is expected to be used by people with different goals and knowledge and where the hyperspace is reasonably big. AH techniques can be extended to time-based media, particularly in the domain of IDTV [6], e.g. dimming an item by showing it in a smaller screen portion when it is not particularly interesting for the viewer, offering
Figure 7. Adaptation rule for Option 2 additional information for the user when he/she has little knowledge about the subject proposed, etc. These ideas should be taken into account in order to properly design the different options for our self-adaptive SCOs. Concerning self-adaptive SCOs, a very close idea is exposed in [7], where requirements on an e-learning standard for adaptivity support are established. ADL SCORM is inspected referring to these requirements, suggesting exemplary enhancements to support adaptive e-learning. These enhancements consist in improving SCOs with several mappings instead of a fixed one. The difference with our proposal is that our selfadaptive SCO has to configure itself according to user’s characteristics, whereas in [7] an ITS should make the selection of the appropriate mapping. Moreover, it does not define a general frame for the SCOs to work in.
7. Conclussions and Future Work In this paper we have presented a solution for personalizing learning objects in the field of t-learning by means of self-adaptive SCOs. This solution consists in extending the SCORM Data Model with a vocabulary of adaptation parameters, that are some user’s characteristics relevant in the context where the learning experience is taking place. In this manner, the SCOs can access information out of the scope of the course they belong to, concerning user’s preferences and background, and configure their behaviour accordingly. Self-adaptive SCOs can show different options of appearance and behaviour, the properties for each option are stored in a configuration file. In order for the SCO to know which configuration file to use in a concrete execution, adaptation rules are provided in the adaptation file. These rules establish relationships between the
values for some adaptation parameters and the different behaviours the SCO can adopt. These SCOs can be manually created by programming the Java template and writing the XML files corresponding to configuration and adaptation files. However, the content creator does not need to know programming or the syntax of XML files. To make this creation easier, we have developed an authoring tool, called SCOCreator, which permits creating self-adaptive SCOs from existing Java templates, without writing the Java template or the XML files manually. One of the most relevant characteristics of our proposal is that it is modular both for individuals and systems. Each individual implied in the process has its own function: the content creator does not need to know the user model of the ITS, whereas the vocabulary creator has to be an expert at the user model of the ITS he/she creates the inference rules for, but does not need to understand how the learning objects are created. Moreover, our solution works for every ITS on reception as long as the vocabulary creator had previously provided it with the inference rules according to its user model. This independence of roles can also be noticed in the design of the SCOCreator tool, since the content creator does not need to program the SCOs but he/she should be an expert of the theme they teach. On the other hand, the Java templates are created by an expert programmer who does not need to know anything about the actual contents of these templates. As a future line of our work, we are putting the final touches to a proposal to provide other components of the SCORM CAM with adaptivity, concretely, activities. This proposal will consist in offering different ways of achieving the intended objectives for an activity, using the same approach as self-adaptive SCOs, extending SCORM metadata with adaptation rules. Moreover, we intend to provide t-MAESTRO with the ability of improving SCORM courses by making them more entertaining adding TV programs related (or segments of them) to these courses. As an example, if t-MAESTRO, which stores the user profile, knows that the user likes football and he/she is following an English course, it can offer him/her a football match in English in order for him/her to improve his/her listening comprehension. In order for the ITS to establish these relationships, two ontologies are needed: an ontology of educational elements based on the SCORM standard [8], to store instances of the SCORM elements the ITS have access and their interrelationships; as well as a TV ontology [3] 2 , to act as a repository of the information related to TV programs.
Acknowledgements Partly supported by the R+D project TSI 2004-03677 (Spanish Ministry of Education and Science) and by the EUREKA ITEA Project PASSEPARTOUT.
References [1] ADL. Sharable Content Object Reference Model (SCORM). http://www.adlnet.org, 2004. [2] Peter J. Bates. A Study into TV-based Interactive Learning to the Home. http://www.pjb.co.uk/t-learning, 2003. [3] Yolanda Blanco-Fernández, José J. Pazos-Arias, Alberto Gil-Solla, Manuel Ramos-Cabrer, Jorge GarcíaDuque, Rebeca P. Díaz-Redondo, Ana Fernández-Vilas, Belén Barragáns-Martínez, and Martín LópezNores. Avatar: Advanced telematic search of audiovisual contents by semantic reasoning. In 4th Work2
http://idtv.det.uvigo.es/es/avatar/ontologia.html.
shop on Personalization in Future TV - Methods, Technologies, Applications for Personalized TV, Eindhoven, The Netherlands, 2004. [4] Peter Brusilovsky. Methods and techniques of adaptive hypermedia. User Modeling and User Adapted Interaction (Special issue on adaptive hypertext and hypermedia), 6(2-3):87–129, 1996. [5] DVB Consortium. Multimedia Home Platform Specification 1.2.1. European Standard ETSI TS 102 812 V1.2.1, 2003. [6] Judith Masthoff and Lyn Pemberton. Adaptive hypermedia for personalized TV. In Adaptable and Adaptive Hypermedia Systems, pages 246–263. IDEA group publishing, 2005. [7] Felix Mödritscher and Victor Manuel García Barros. Enhancement of SCORM to support adaptive ELearning within the Scope of the Research Project AdeLE. In Proceedings of the ELEARN 2004 Conference, Washington, USA, 2004. [8] Marta Rey-López, Rebeca P. Díaz-Redondo, Ana Fernández-Vilas, and José J. Pazos-Arias. Entercation experiences: Engaging viewers in education through tv programs. In 4th European Conference on Interactive Television (EuroITV 2006), Athens, Greece, may 2006.
