Project Ubi-Learn: an Intermediation Infrastructure for Multi-channel Accesses to Future LMS Vincent Chevrin
Alain Derycke
José Rouillard
Laboratoire Trigone Université de Lille 1
[email protected] 59655 Villeneuve d’Ascq cedex France
Laboratoire Trigone Université de Lille 1
[email protected] 59655 Villeneuve d’Ascq cedex France
Laboratoire Trigone Université de Lille 1
[email protected] 59655 Villeneuve d’Ascq cedex France
Abstract Using previous experiences of designing elearning systems, we have started a new project, UbiLearn, which aims to support the remote learners in various situations, mobile or fixed. This means a great variety of communication channels such as Web, email, SMS, phone call… This leads us to consider a large range of scenarios of use, where channels are combined in a same unit of interaction depending of the context. In the context we capture not only information about the technological and physical contexts of interaction, but also the user profile, and the nature of the activity and its history. Ubi-Learn intermediation is based on the assumption that the learning system is factorised into a collection of dedicated e-Services, and that the user interactions with the channels can be abstracted with a channel independent modelling language. The management of these intermediations is achieved through a multiagents infrastructure combining several sources of contextual knowledge, especially for the dynamic orchestration of the e-Services.
1. Introduction In the context of our past researches, we had drawn lessons from previous design and operations of technological platforms, such as Learning management Systems, LMS, including support for LearnFlow design and execution, and various portals and tools for Computer Supported Collaborative Learning [1]. We have also noticed the raise of new needs for future distance and open learning organisations: development of learner practices and attitudes about mobile communication (both communicational and computing activities), the needs for an openness of the future systems in terms of interoperable cross-organisation learning systems…. In this framework we have started
the design and implementation of a new infrastructure for the intermediation between the learners, with fixed or mobile locations, accessing the system through various devices and value-added networks, and a collection of learning services, which can evolve and be extensible, depending the context. This new infrastructure is the base of our project Ubi-Learn that aims to support more mobile and flexible learning processes by integrating recent progresses and development of both the software engineering domain, especially the Service Orientation, and of the Pervasive Communication and of the context-aware computing domains of research. The problem of the intermediation in our case is complex, due to the number of settings that can be encountered, and to the needs of flexibility and dynamic adaptations. This implies that some decisions about the composition (of medium, channels, services, etc.) must be taken at run-time and not at design-time. So our technological infrastructure is designed with these constraints, and provides capabilities to be extend to support new communications channels or new user devices in the future. Of course it is relatively too early to assess this infrastructure in real situation of a distance learning process. The test of an infrastructure in the perspective of a user-centred design is still a difficult problem [8]. However in order to test some assumptions and subparts of the infrastructure, we have already conducted, in parallel, limited evaluations using dedicated prototypes, for example to evaluate usability and efficiency of multimodal interactions [7], combining Web browsing and voice channel [15]. In the present paper, we will focus on the design and implementation of the infrastructure and on the different technologies that we have mobilised for this purpose.
Proceedings of the Advanced International Conference on Telecommunications and International Conference on Internet and Web Applications and Services (AICT/ICIW 2006) 0-7695-2522-9/06 $20.00 © 2006 IEEE
2. Complexity of multi-channel intermediation in distance and open learning 2.1 Diversity of the scenario of use We can start with a rapid analysis of some transactional scenarios in the field of e-Learning systems, in the context of future mobile or ubiquitous learning. We proposed three typical scenarios of use: S1: The learner follows a learning process through an interaction with an LMS, using her/his traditional Web browser. Encountering a problem with his/her progression, he/she contacts the assigned tutor through a phone call, via a dedicated call-centre. S2: The learner interacts with a pedagogical object, or a specific learning e-Service. Interaction combines in a same unit of time (synergetic mode) two modalities: graphical through the mouse and the screen, vocal through a voice recognition and a voice synthesis dialogue system. S3: Is a variant of S1, where, when the learner is in contact by phone (for example using a voice over IP channel) she/he can simultaneously navigate on the Web, in synchronous collaboration with her/his tutor, though a co-browsing mechanism and set-up ( Web proxy). S1 corresponds to a multi-channel approach, where the two channels used are loosely coupled, both at the semantic level and at the temporal level, because they can be simultaneous or at the opposite very asynchronous. S2 corresponds to what it is called Multimodality in the field of Human Computer Interaction. S3 is hybrid and can shift from a multichannel approach to a multimodal one, depending of the semantic and temporal coupling between the channels, and of the nature of the tutor agent: human or a synthetic counsellor. Our aim is to support the two kinds of interaction inside a same system, with fixed or mobile connections. From these scenarios it appears clearly a great variety of possible combinations of potential channels such as Web, email, SMS, Phone call…, used exclusively or in conjunction, synchronously or not, in order to support the different phases of the transaction between the learners and the services. Whatever these services are human, such as tutoring, or computerized such as the LearnFlow management, digital libraries, ePortfolio, etc. The channels are defined, for us, as a particular association of one of user’s devices (UDs: workstation, PDAs, smartphone, etc.) and a particular access network: for example Wifi on a campus wide Intranet,
ADSL link at home, or mobile-phone networks with GRPS data link. The heterogeneity produces a great complexity due to all the potential combination and coupling, tightly or loosely, of these channels, and of their possible interference. In order to cope with this complexity we have derived a theoretical framework [6] which able us to do some predictions about the communicational properties of the channels used, inspired from the Grounding concept in psycholinguistic, for example for the temporality of communication modes, and of the Task-Media Fitness theories. This is also used to fix the good match between the nature of the learning task, and its documents, and the appropriate channel to deliver them. 2.2 What is a channel: a combination of user’s devices and networks? We have derived a general ontological model of the intermediation, for our case, and for the capture and evolution of the contexts in several sources of knowledge. This has been done considering the literature about ontological models for context-aware computing, such as [4] [9], enriched with the studies of proposal coming from standardization bodies (for example W3C consortium). Roughly the sources of knowledge, to support context-aware interaction, represent four domains: 1) Knowledge about the e-services, Activities supported, Roles, and Policies issued by the supporting learning organisation. This model emphasis the knowledge provided by a meta-model of the human activities based on our previous works in the field of the CSCL using the Theory of Activity as a foundation [1] [2]. This means that this meta-model is used as knowledge to adapt the services to a particular learner and past history, especially to his/her Role(s) into the running activity. This is closed to the meta-models of some Educational Modelling Languages such as IMSLD, aiming to support pedagogical scenarios; 2) Knowledge about the user (profile, technical and physical contexts) in the vein of CC/PP; 3) Knowledge about the nature of the channels, theirs coupling capacities and theirs constraints; 4) Knowledge about the intermediation itself (rules for the use of channels, personalisation of interaction, management of the interaction histories…) derived from our aforementioned theoretical model. At this stage our effort in term of modelling and capture of context knowledge is put mainly on the first and fourth ones. For a first step in our design we don’t consider sophisticated mechanisms, based for example on automatic machine learning processes, to produce
Proceedings of the Advanced International Conference on Telecommunications and International Conference on Internet and Web Applications and Services (AICT/ICIW 2006) 0-7695-2522-9/06 $20.00 © 2006 IEEE
new knowledge about the contexts. But the technologies we have selected to build our infrastructure, I.E. the use of a Multi-Agent System (MAS), are compatible with this potential extension.
3. The goals and requirements for the Ubi-Learn intermediation platform 3.1 The technological framework of Ubi-Learn derived from works on Ubiquitous Computing Our project reuses previous researches done in the field on context-aware computing or ubiquitous computing. We derived from its some aspects of the context modelling and management. We take a special attention to the works done around devices independences interaction such as promote by the W3C consortium, and on the ways to do dynamic adaptation of the document contents and of user interface elements in accordance with these devices and channels. For this purpose we reuse both our experience in the design of Web server coupled with speech interaction (ad hoc prototypes) using our expertise in VoiceXML technology [7], and an adaptation of our abstract language, called PlasticML [15] to specify interaction and to support generation of the concrete interface through a multistage XSL-T transformation of the document and elements for interaction with the user. This approach offers a potential to support various User Devices and various presentation languages: XHTML, cHTML, WML, VoiceXML… 3.2 A Service-Oriented Approach of the mobilelearning Following the movement already started in the field of software engineering, with the Service-Oriented Architecture (SOA) and in the business field [16], we can show that the Service-Oriented Architecture is also a good framework to reframe the Learning and Content Management Systems. This is confirmed by proposals for the future learning systems, seen as a collection of e-services, such as those of Wilson et al [20]. If we analyse these proposals we can see that the architecture of services is decomposed into three layers from bottom to top: the underlying layer concerns the data and their persistence across the services relying on some databases. The intermediate layer provides application services (for education it is for example LearningFlow, e-portfolio, authoring, etc.) and common services (AV conferencing, role management, DRM…). The upper layer not described precisely in [20] is called User Agents and is in charge of managing the interaction between the user (through his/her DUs) and a collection of services that
composed his/her applications. This hypothesis about the importance of these future open frameworks of learning services in one of the main assumptions behind the design of the Ubi-Learn infrastructure, with a focus on the upper layer. From the technological viewpoint this can be achieved through the Web-Services technologies, described by the triple: “WSDL”, “UDDI” and “SOAP”; and others proposals [14]. However these technologies exhibit some limits, especially for dynamic or opportunistic composition or orchestration of the services that we must fix in our infrastructure.
4. The general architecture of Ubi-Learn 4.1 The Channel-Service relationships We will focus, now, more precisely on the nature of the problem of associated the channels used for interaction with the services. From different scenarios we have shown than the composing of the different e-services (corresponding to different learning tasks), their interleaving, are conditioned by the nature of the channel and their interaction styles (direct manipulation versus speech dialogue). Effectively from our previous evaluation of first prototypes [7], combining voice and Web or Wap interaction modalities, we have discovered that the concurrent accesses to several e-services conducted to different interleaving of the e-services primitives, depending of the channel used. For example the reading and the choice into a table of items are different for voice interaction (because we cannot force the user to remember all the lines content) than for Web interaction, where it is possible to see the whole table on the screen. It means that the fragmentation of the source document (for example a form) must be adapted to the channel characteristics, same for the interaction elements. 4.2 A specific middleware for managing the intermediation between the learners and the services From the above requirements and analysis we have based the overall design on a middleware specific for intermediation between communication channels and the e-Services. It provides “adaptators” of all the elements connected to this middleware and transformation engines, guided by rules, in order to transform the flow of documents and interaction elements in input and output. This can compare to two others specific middleware for that purpose [11] for the field of military application, or those of [18] for the mobilelearning domain. The requirements are: 1) dynamicity
Proceedings of the Advanced International Conference on Telecommunications and International Conference on Internet and Web Applications and Services (AICT/ICIW 2006) 0-7695-2522-9/06 $20.00 © 2006 IEEE
of e-services orchestration; 2) dynamicity of the coupling of channels, either to support a multi-channel or multimodal approaches; 3) the needs for persistence and management of the rupture [7] in the transactions;4) and adaptation to knowledge sources (see figure 2). This has convinced us that a multiagents technology will be the best for us. This is of course confirmed by the raising of more and more applications of these multiagents systems for example in the field of context-aware computing, management of mobile communication, e-Commerce, orchestration of Web services and workflows…
and communication mechanisms that are compliant with the FIPA standard proposal. This means than in the future the communication will be easier between the Ubi-Learn infrastructure and other FIPA MAS developed systems, for example dedicated to the building by inference of some context knowledge. This is one of reasons, combined with an e-services approach, of the openness of our proposal. The figure 1 gives the agents and their role in a transaction for a specific learner. Only agents directly in charge of documents and interaction processing are represented. Figure 1: Agents and their roles in a transaction with a specific learner
5. The technological implementation of Ubi-learn
LMS IN SOA
5.1 The dynamic coupling of channels and modalities The three contrasted scenarios of use presented in section 2.1, illustrate the needs for mono-channel, multi-channel, and multimodal type of interactions, and show how it can necessary to shift smoothly from one mode to another one, using dynamic association of channels in input and in output. This is solved for the coupling of a traditional Web-server (visual interaction) with and VoiceXML one, able to manage speech dialogue interaction through the web navigator or the telephone (mobile or not) [7] [15]. In input it means that is possible to merge some inputs, after their transformation to a same format, provided by the PlasticML specification and engine, in order to construct a request to a particular service. In output it is the possibilities, on the fly, to transform information provided by the services, into the good format (same done by PlasticML engine [15]), or even to split it toward two different channels (ex: voice + SMS). 5.2 The Multi-agents technology selected, JADE, for the implementation of Ubi-Learn We can now present and justify the use of a MultiAgents system (MAS), JADE, (http://jade.tilab.com), as the underlying layer for the implementation of the Ubi-Learn middleware. JADE is provided by a consortium of companies especially from the telecommunication field, and it has been used in the design and deployment of large projects implying use of mobile devices. It is possible even to push an agent inside a Smartphone. It has already been used successfully either for the dynamic orchestration of Web-Services, or the processing of the Human Computer Interaction for both multimodality and dynamic adaptation of contents to the user devices. Another argument for the use of JADE, as our MAS, is the underlying technologies that are Java programming
Learner i
E-SI
PHA
ESA
CAA
RAA
PIA
MM CA
E-SI TMA
CAA
Channels
Networks Devices
Definition of the main User Agents (UA) and other components for a transaction Ti: eSI: e-Services Interactive
ESA: e-Service agent
PIA: Personalization Interface Agent
MMCA: Multi-channel & Multimodal Communication Agent
CAA: Channel Adaptation Agent
TMA: Transaction Management Agent (pour Ti)
RAA: Role Activity Agent
PHA: Persistence Historical Agent (for learner I) flows of documents and interaction elements
5.3 The problem of a multi-channel and multimodal adaptation of learning objects in the SCORM standard We can give a special emphasis on the problem of the dynamic adaptation of the user interface, in accordance with the characteristic of the channel used, if the “document” provided by an e-Service is a Learning Object (LO) compliant with the SORM standard. It appears that some adaptations, for example towards multimodal (speech + visual) interaction, are not totally satisfying due to constraints of the SCORM model of interaction, and we can propose improvements. For this purpose we have conducted a case study [3] of the transformation of SCORM LO for non Web compliant user platform: for the WAP platform on mobile phone, and voice interaction for phone, fixed with Voice-Over-IP connection, or mobile through the GSM. We have no place here to present into details our investigations, but it appears that the choices done for the establishment of the SCORM standard, prevent the reuse of these LO into the mobile and context-aware learning modes. The problem occurs at different levels: 1) The choice of a specification based directly on HTML format with some technical “tricks” such as using a frame of dimension zero, “invisible” for the
Proceedings of the Advanced International Conference on Telecommunications and International Conference on Internet and Web Applications and Services (AICT/ICIW 2006) 0-7695-2522-9/06 $20.00 © 2006 IEEE
user”, in order to attach specific scripts for communications between the LO and the LMS platform; 2) The choice of using the JavaScript language to support not only user interaction but also data exchanges, such as score, with the LMS. It must be known that the script languages, into the two cases we have mentioned, are not compatible. It is WML script for WAP and ECMA script for VoiceXML. The limit of the SCORM specification, regarding the reusability in different contexts of interaction, has been already pointed by other authors: for example [10] show that, even if it is used only in the Web mode, it can result a mosaic effect of the composition on the same screen of two or more SCORM LOs; Other [17] proposed to change the specification of SCORM using an XML format for the document in order to avoid this effect. Whatever none of these critics of SCORM standard point the problem of its deficiencies, relative to the future pervasive computing and communication contexts. Our solution have been first to abandon partially the compliance with SCORM by embedded into the Learning Object a description done in our PlasticML language, and second to modify a SCORM compliant, open source, LMS platform, for example www.Claroline.net, in order to support interaction with these new LOs through different access channels. But this reengineering of current LMS shows still difficulties and lacks of flexibility to support ubiquitous interaction. This is why we are convinced that our approach of a more general infrastructure will be more powerful, even if it required more effort for its design and implementation.
6. Conclusions The figure 2 gives an overview of the Ubi-Learn project when all the knowledge sources, described in section 2.2, will be implemented and connected to our intermediation middleware. Of course most of these context-aware services would not be designed and implemented only by us. In effect it appears that more and more researches project are focused on open and reusable context services [12] that could be coupled with our infrastructure. Same it appears that some proposals, done by the telecom operator such as those of Liberty Alliance, especially in the framework of the future Intelligent Network, will provide sources of new information on the context. It must be noted that all these external context-aware services are accessible through the Web Services technologies and could be
easily interfaced with our Ubi-Learn infrastructure, thanks to SOA. All the agents, as described in the figure 2, and the management and deployment tools for this agent population, not represented in this figure, are operational now. We have connected the intermediation middleware on a limited number of learning e-services especially designed for this purpose. We are waiting for the future implementation, in open source software, of the services announced in the ELF framework of [20], in order to enrich the functionalities. The ongoing work is done into two directions: - A better use of the potential of a software agent approach to the dynamic orchestration of the e-Services, in the context of the user activity (profile for his/her self-service), and of his/her present Roles into the activities. - The improvement of the multimodality interaction by reintroduced our results of parallel projects [6] [7], which have been experimented and evaluated more deeply. Effectively we are convinced, especially in the framework of MobileLearning that the use of the multimodality will be effective and an improvement of the interaction with the learners.
7. References [1] Bourguin, G., Derycke, A, “A Reflective CSCL Environment with Foundations Based on the Activity Theory”, Springer Verlag proceedings of ITS’2000, Fifth International Conference on Intelligent Tutoring Systems, Montreal, CANADA, 19-23 June 2000, LNCS vol. 1839, 272-281. [2] Bourguin, G. Derycke, A, “Co-evolution inside interactive systems: Reflections about results coming from the Activity Theory for supporting Collective and Distributed Practices”. In RIHM, Revue d’Intéraction Homme-Machine, vol. 6, n°1, June 2005, 30 pages (in French). [3] Bouyer, A. Plasticité des interfaces Homme-Machine et Objets Pédagogiques structurés. MS thesis in Computer Sciences, Université de Lille, France (in French) June 2004, 52 pp. [4] Chen, H. Finin, T. Joshi, A. Kagal, L. Perich, F. Chakraborty, D, “Intelligent Agents meet the Semantic Web in Smart Spaces”. In IEEE Internet Computing, NovemberDecember 2004, pp 69, 79. [5] Chevrin, V., Derycke, A., Rouillard, J, “Some issues for the Modelling of Interactive E-Services from the Customer Multi-Channel Interaction Perspectives”, IEEE international conference on e-Technology, e-Commerce and e-Service, IEEE Press, Hong Kong, pp. 256-259, April 2005.
Proceedings of the Advanced International Conference on Telecommunications and International Conference on Internet and Web Applications and Services (AICT/ICIW 2006) 0-7695-2522-9/06 $20.00 © 2006 IEEE
[6] Chevrin, V., Rouillard, J., Derycke, A, “Multi-channel and multi-modal interactions in E-marketing: Toward a generic architecture for integration and experimentation”. HCII international conference, Las Vegas, 22- 227 June 2005, Lawrence Erlbaum editors, 10 pages. [7] Derycke, A., Rouillard, J., Chevrin, V., Bayart, Y, “When Marketing meets HCI: Multi-channel customer relationships and multimodality in the personalization perspective”. HCI International 2003, Heraklion, Crete, Greece, 2003, Lawrence Erlbaum editors, Volume 2, pp. 626-603. [8] Edwards, K. Belloti, V. Dey, A. K. Newman, M, “Stuck in the middle: the challenges of user-centred design and evaluation of the infrastructure”, April, 5-10 2003, Ft Lauderdale, Florida, ACM Press, pp 297-304. [9] Henricksen, K. Indulska, J. Rakotonirainy, A. “Modeling Context information in Pervasive Computing Systems”. In proceedings of 1st International conference of Pervasive Computing, January 2003, Springer Verlag, LCNS vol 2414, pp 167-180. [10] Ip, A., Radford, A. and Canale, R, “Overcoming the presentation Mosaic Effect of Multi-Use Sharable Content Objects”. Proceedings of the 20th Annual Conference of the Australasian Society for Computers in Learning in Tertiary Education ASCILITE''03 (7-10th Décembre 2003, Adelaide, Australie), ASCILITE, 2003, pp. 256 – 26 [11] Lard, J. Segdobo, C. Bisson, P, “Advances in software architecture design applied to human computer interaction processing”. SIM’04 Semantic intelligent Middleware for the Web and the Grid, a workshop of ECAI conference, Valencia, Spaign, 23-27 August 2004. [12] McFadden, T. Henricksen, K, Indulska, J. Mascaro, P, “Applying a Disciplined Approach to the Development of a Context-Aware Communication Application”. Proceeding of the 3nd IEEE int’l Conf. on Pervasive Computing and Communications (PERCOM 2005), 7p.
[13] Oviatt, S, “User-centered modeling and evaluation of multimodal interfaces”. Proceedings of the IEEE, Volume: 91, Issue: 9, Sept. 2003. Pages: 1457-1468. [14] Papazoglou, M. Web Services and Business Transactions. World Wide Web and Web Information Systems, 6, 2003, Kluwer academic press, pp49-91. [15] [Rouillard, J, “Plastic ML and its toolkit”, HCI International 2003. Heraklion, Crete, Greece, 2003. Lawrence Erlbaum associates, publishers. Volume 4. Pages: 612-616. [16] Rust, R. Kannan, P.K, “E-service: a New Paradigm for Business in the Electronic Environment”. Communication of the ACM, June 2003, vol. 46, N°6, pp 37-42. [17] St-Pierre, R. and Hope, P, “Dynamic Appearance Model Analysis and Alternatives”. Canadian Department of National Defence, Published by Sun Microsystems Inc, 2003. (http://training.sun.com/ US/images/pdf/DynamicAnalysis.pdf) [18] Trifonova, A. M. Ronchetti, “A general architecture for Mobile Learning”. In Proceeding of the ICALT 2004 conference, IEEE press. [19] Tsalgatidou, A. Pilioura, T. “An Overview of Standards and Related Technology in Web Services”, Distributed and Parallel Databases, 12, 2002, Kluwer academic press, pp 135-162. [20] Wilson, S. Blinco, K. and Rehak, D, “Service-Oriented Frameworks: Modeling the infrastructure for the next generation of e-Learning Systems” Paper for the Alt-I-Lab 2004 conference (July, 2004)
Others basic services and partners Learning e-Services interactive
Learner population
Abstracts Interfaces concretes Interfaces
Channels and Networks Ubi-Learn Infrastructure
Agents Adaptation to e-Services & management Adaptation to channel compositions & Devices Adaptation to the Activity and Context Supervision and management of N transactions
Definitions eSI
Compos. rules
comm. rules
Roles, activiies
usersProfiles
Channel Models
Knowledge Bases
Figure 2: an overview of the Ubi-learn general architecture
Proceedings of the Advanced International Conference on Telecommunications and International Conference on Internet and Web Applications and Services (AICT/ICIW 2006) 0-7695-2522-9/06 $20.00 © 2006 IEEE
