24 pages

A knowledge-based framework for e-learning in heterogeneous pervasive environments

of 24
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
ABSTRACT We propose a ubiquitous learning approach useful not only to acquire knowledge in the traditional educational meaning, but also to solve cross-environment everyday problems. By formalizing user request and profile through logic-based
  A knowledge-based framework fore-learning in heterogeneouspervasive environments   Michele Ruta, Floriano Scioscia, Simona Colucci, Eugenio Di Sciascio, Tommaso DiNoia, Agnese Pinto Politecnico di Bari, via Re David 200, I-70125 Bari, Italy{m.ruta, f.scioscia, s.colucci, disciascio, t.dinoia}@poliba.itagnese.pinto@doom-srl.it ABSTRACT We propose a ubiquitous learning approach useful not only to acquire knowledge in thetraditional educational meaning, but also to solve cross-environment everyday problems.By formalizing user request and profile through logic-based knowledge representationlanguages, a lightweight but semantically meaningful matchmaking process is executed in orderto retrieve the most suitable learning resources. Standard formats for distribution of learningobjects is extended in a backward-compatible way to support semantic annotations in ourframework.Framework and algorithms are absolutely general purpose, nevertheless an application hasbeen developed where the semantic-based Bluetooth/RFID discovery protocols devised inprevious work, support users –equipped with an handheld device– to discover in the environmentlearning objects for satisfying their needs. INTRODUCTION Pervasive e-learning has been investigated in recent research because of its evolutionary impacton the definition of traditional e-learning: learning anytime, anyhow and anywhere. The maingoal is to take full advantage from the possibility of performing the knowledge acquisitionprocess also in case of lack of fixed infrastructures. Many studies recognize the independence of the learner's physical location and the availability of powerful learning devices as the main addedvalue of electronic learning with respect to traditional approaches (Maurer, 1998). Hence, a fullexploitation of ubiquitous computing technologies can deeply affect the most significant aspectsof e-learning systems. The main issues of the so-called mobile learning (m-learning) are identified and gathered in(Sharples, 2007). Beyond achieving benefits of electronic learning, m-learning allows a highercustomization of the learning experience through adaptive techniques for content provisioningand organization. From this point of view, it is important to combine the usefulness of both e-learning approach and mobility technologies within a unified vision. Pervasive and Web-based  technologies should be applied together in defining frameworks and guidelines to really allow auser to learn anywhere she is. The main challenge –or opportunity, we daresay– is to enable theknowledge acquisition across contexts and environments, rather than simply exploiting handhelddevices for the fruition of learning contents. Hence, there is the need to move away from“adapting” activities and approaches designed for personal computers to mobile devices andcontexts. On the contrary, a comprehensive approach should be outlined, taking into account: •   the complexity of mobile scenarios: the benefit of learning ubiquitously by using a portabledevice is balanced by the technological constraints of such devices (limited memorycapacity, reduced computational capabilities, restricted battery power, small screen size,among others); •   the different dialectic relationship learners establish in those contexts with respect to wiredones.Flexible and context-aware discovery techniques thus become a key element to buildpervasive learning infrastructures allowing a great personalization according to individualrequirements, possibilities and context, also coping with the high differentiation of current mobiledevices.In spite of the growth in the diffusion of wireless-enabled handheld devices providing thenecessary connectivity for complex applications, in general they are based on short range, lowpower technologies like Bluetooth (Bluetooth), which grant a limited interaction among hosts.Furthermore, as ubiquitous contexts are very volatile environments, some important issues arestill open. Particularly, services or resources are often unavailable because the location of mobileproviders can change unpredictably (Chakraborty et al., 2001). Hence, an advanced discovery of learning resources has to be flexible and decentralized, to overcome difficulties due to the hostmobility.We borrow languages and technologies from the Semantic Web vision and adapt them topervasive contexts in order to produce a framework fully interoperable with fixed approaches aswell as with accepted standards for learning contents modeling. In this paper, a coherentknowledge-based retrieval of mobile learning resources has been devised and implemented.Resources are advertised over a mobile ad-hoc environment as learning objects according to theLOM –Learning Object Metadata– standard (IEEE, 2002), supported by SCORM –SharableContent Object Reference Model– specification (SCORM, 2004) for Learning ManagementSystems (LMS). The learning content needs to be redesigned to meet the requirements of a mobile exploitation(Keil-Slawik et al., 2005): in our approach learning resources have a formal characterization.Independently on the chosen syntax, learning modules are modeled as Learning Objects (LOs)according to LOM. But we propose to extend the standard to provide a semantic annotationunambiguously describing the learning object with respect to a specified ontology. The context surrounding the learner is modeled by exploiting LIP –Learner InformationPackaging– standard (LIP, 2001). Also in this case we extend LIP specification to deal with thesemantic annotation of learner contextual information.Given a learning need (user request), LOs are automatically retrieved following semantics of their descriptions. Furthermore they are ranked according to the degree of correspondence withthe request. Both learning needs and resources have to be conveyed through annotations in OWL-DL (W3C, 2004). It is a formal representation language based on Description Logics (DLs)formalism (Baader et al., 2002), which allows interoperability with Semantic Web technologies  and also enables a set of reasoning services. Abduction and contraction algorithms presented in(Di Noia et al., 2007) have been adapted for being performed by a mobile device.A learner-centric perspective is adopted, providing expertise on demand solutions for self-training, i.e. , supporting a pull model for learning resource discovery and acquisition. Ourapproach is general and protocol-independent. Nevertheless it has been motivated in a cross-environment learning scenario where the semantic-enhanced Bluetooth/RFID discovery protocolspresented in (Ruta et al., 2007b; Ruta et al., 2007a) are exploited as underlying interactionparadigms. The paper is structured as follows. In the next section a motivating scenario should clarify theapproach and the rationale behind it whereas in further section the proposed framework andalgorithms are outlined. Finally, we comment on related work before conclusions. MOTIVATING SCENARIO Ubiquity, universality and efficiency are the main requirements for a knowledge-basedframework aiming at supporting highly relevant and context-aware discovery and sharing of learning resources for self-training. In particular, our goal is to provide enough flexibility tosupport knowledge acquisition in informal and unstructured settings, in addition to moretraditional and structured ones ( e.g. real or virtual classrooms). This kind of use case can clearlyshow the benefits of adopting ubiquitous computing technologies in a multi-agent framework forgoal-oriented knowledge acquisition and learning. The approach and algorithms are basicallyhardware and O.S. independent. Equipment features are taken into account when it comes toselect best matching available learning resources.Bluetooth technology is increasingly adopted in a variety of devices and appliances beyonddesktop and mobile computers. This could allow exploitation of semantic-enhanced Bluetoothresource discovery protocol (Ruta et al., 2006) in many different contexts, in order to findlearning modules matching with user’s interests, needs and constraints. Pervasiveness is increasedby embedding semantic-enhanced RFID technology into an environment (Ruta et al., 2007a).Objects can self-describe to nearby RFID-enabled mobile devices through their attached RFIDtag, so becoming knowledge resources for helping the user to perform her intended task. The proposed approach is described and motivated referring to a scenario outlining learningneeds occurring to a woman,  Janet , in typical daily activities. In the morning, Janet is driving her newly purchased car to her workplace. She is stillunfamiliar with advanced car controls. In particular, she is currently wondering how to store astation within the memory of the car radio system. She uses her Bluetooth-enabled mobile phoneto discover such an information from learning modules supplied by her car’s computer.  The car computer exposes the topics of the manual which can be discovered via the semantic-enhanced Bluetooth Service Discovery Protocol. Each topic is packaged as an atomic learningmodule, but dependencies and references between modules can be present. Each learning moduleis described by means of a semantic-based expression of its content and requirements for fruition. The mobile semantic matchmaker installed on the mobile phone could then perform a discoveryprocess to find the learning resources best fitting the user’s request and profile. Both areexpressed in a reference ontology-based formalism in order to be matched with available LOs(whose semantic characterization follows the standards-compatible format extension outlinedlater on).It is important to point out the differences between user request and profile. The requestexpresses the learning needs and goals of a user, whereas the profile describes her current context  in terms of: background knowledge and training; time and place constraints; technologicalrestrictions imposed by software/hardware features of the user device. Hence, the request varieswith each knowledge discovery process. The envisioned framework should support applications with both explicit and implicit userinteraction paradigms. In the former case, a request is directly composed by the user andsubmitted to the embedded mobile matchmaking engine. In the latter case, the user implicitlytriggers a support request by performing a particular interaction with elements of a smartpervasive computing environment. The request is then built in a semi-automatic way, by interpreting the current user action andformalizing her intention into an information/knowledge need, while possibly leaving room fordirect customization. On the contrary, the user profile changes with less frequency and generallyin an automatic fashion, e.g. by updating the description of user location, characteristics of herdevice and the knowledge and experience she has gained.User request and profile have their counterparts in the annotation of a learning module, in theform of description and requirements respectively. The description expresses the topics andcontents of a learning resource in an unambiguous way, according to a reference ontology whichmodels a broader discipline. On the other hand, requirements model necessary conditions foradequate fruition and comprehension of a learning module. They can concern (but are not limitedto): (a) prerequisites on cultural or technical background knowledge; (b) time and locationconstraints for learning module fruition ( e.g. a silent room is needed for an interactivepronunciation lesson); (c) constraints on hardware/software features for accessing a learningmodule ( e.g. playing videos in a particular format with a certain minimum screen resolution). In amatchmaking session, both elements have to be taken into account. First of all, user profile mustsatisfy the prerequisites for fruition of a LO, otherwise knowledge acquisition cannot occur.Subsequently, the best matching descriptions with respect to user request are computed amongavailable learning objects whose prerequisites are satisfied.As a very small example, Janet’s request can be stated as: user instructions on radio memorymanagement for car sound system. At the same time, her profile can be modeled as: 5 minutes of available time for resource fruition, 240x320 pixel screen and support for Java ME and Flash Liteformats. Let us suppose that – among others – the following user manual topics are provided aslearning objects:A 1 : user instructions on radio memory management for Acme car sound system; length of activity is 2 minutes and format is Flash Lite.A 2 : user instructions on CD player for Acme car sound system; length of activity is 4 minutesand format is Flash Lite.A 3 : user instructions on air conditioning regulation: length of activity is 7 minutes and formatis Flash Lite. The detail level of descriptions reflects the “density” of learning resources for a given domain.In the previous example, single functionalities of station memory management (add, delete,modify) could be explained either within the same learning module or in separate ones.Furthermore, a car manufacturer could provide different sets of learning objects targeted to usersand car electrician respectively. They would be annotated with respect to different referenceontologies. Hence, a requester could limit her search to the desired category through apreliminary ontology agreement with a provider of learning objects.  Janet reaches her workplace at a law firm. She looks in her office library for a book onmaritime law for the case she is currently working on. She does not find it, then she remembers  that her colleague Mark has taken it. Unfortunately, he is away for a meeting. Janet uses hermobile phone to search for learning resources on maritime law which are either offered bycolleagues in her vicinity or available in the knowledge base (virtual library) of the firm. Learning objects and request are modeled in a similar way as in the previous use case, hencedetails are omitted for the sake of conciseness. They refer to an ontology modeling knowledge inthe field of law. Differently from the previous case, the mobile device of the requester collectsresources from multiple nodes. Mobile devices of co-workers are involved, as well as a Bluetoothzone server of the office, acting as a gateway toward learning material owned by the company. This use case is more similar to traditional e-learning approaches, which enable collaboration andresource sharing among learners, as well as access to a central repository of learning resources.Extensions of this use case may include semantic-based composition of learning objects toachieve an on-the-fly mobile courseware definition, given the background knowledge and thecurrent learning needs of the user.  In the evening, after work, Janet goes to the city museum to see an exhibition of Renaissancepaintings. In particular, she would like to learn more about portraits. Paintings are tagged withRFID transponders for both inventory and semantic-enhanced knowledge provisioning tovisitors. A semantically annotated learning object is directly associated to each painting via its RFIDtransponder. Janet’s mobile phone comprises an RFID reader, which can access objectdescriptions. Let us suppose that the following works are currently in her radio range:C 1 : Renaissance oil painting with religious subject;C 2 : Renaissance oil self-portrait;C 3 : Renaissance tempera painting with mythological subject. The mobile matchmaker embedded in the user’s device matches her request with retrievedannotations and results are returned as a ranked list of learning objects. The user can select theones she is interested in and access them. Otherwise, if she is not satisfied with results, she canstart a Bluetooth-based discovery session by interrogating devices of other visitors in the sameroom. With respect to the previous use case, it can be noticed that no central repository of learning resources is present. A “virtual” knowledge base is built instead in a dynamic way, byaccessing relevant content ( i.e. referring to the same domain ontology as the request) providedthrough RFID. Moreover, a peer-to-peer user community is formed in an ad-hoc fashion in orderto exchange learning resources. Such wireless ad-hoc network can be leveraged to satisfy furtherlearning needs, beyond the original one. In the above example, after accessing the learningmodule about self-portrait C 2 (which is the best match for user request), Janet could search forfurther information about portrait as a genre across history. Even though that is not the subject of the exhibition, other visitors –who are supposedly interested in art as much as Janet– mightpossess and share learning material on such topic. User profiles and learning object requirementsare involved into matchmaking, as already explained. This final use case shows the benefits of extending and integrating current smart identification technologies in an overall framework forknowledge discovery and mobile e-learning.   PROPOSED FRAMEWORK In what follows we present a lightweight, hardware and O.S. independent framework to assisteverywhere a learner equipped with a mobile device and wireless technologies like Bluetooth andRFID. The following subsections respectively outlines the discovery architecture, the modeling
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!