<<

Concept-Based System Design to Personalize E-learning

Chao Boon Teo, Robert Kheng Leng Gay Information Institute of Singapore School of Electrical and Electronic Nanyang Technological University Singapore 639798

Abstract: - E-learning aims to foster significant learning improvements through the use of advanced and proven educational techniques. It encompasses the vision of achieving the essential educational goal of lifelong learning. However, literature reviews on the current e- learning system identified some significant shortcomings. Among these shortcomings, we single out the lack of personalization, collaborations and the failure to adopt technological advancements to enhance learning as the key impediments to successful adoption of e-learning. Integrating our understanding across the surveyed literature, we proposed a concept-based knowledge environment that is based on modern like Semantic Web, Grid Computing and Web Services.

Key-words: - e-learning, concept map, distance learning, instructional design, learning objects

Existing research tends to emphasize on the 1. Introduction standardization of metadata standards and the structure of learning resources. Although these are E-learning aims to foster significant learning challenging intentions to enable widespread improvements using the most advanced and adoption, these challenges are not the only proven educational techniques. The advancement important aspect within the comprehensive vision. of and the advent of the information This paper identified the lack of highway coupled with e-learning have eliminated personalization, collaborations and adoption of barriers of time, distance and socioeconomic technological advancements for learning as the status, creating a whole new dimension to key impediments to the successful adoption of e- learning. The dramatic opening of access to the learning. Therefore, to address these limitations, a ever burgeoning deposit of knowledge has concept-based knowledge environment for increased the need for collaborative learning. interactive e-learning was designed to exploit However, despite its ambitious vision, modern technologies to provide the literature reviews unveiled major concerns personalization of learning experiences. This regarding its effectiveness and appropriateness [1, environment employs concept map educational 2]. Although there are many strands of research, advantages to build up knowledge in our no person or organization is solely responsible for collaborative e-learning system. Each concept will its development or standardization. With no function as digital objects to serve as the most validating guidelines, organizations are jumping basic form of learning. into the arena and proclaiming their products as This paper is organized as follows: Section 2 ‘e-learning compliant’. However, what they are presents the skeleton system architecture and key doing is just purely automating their services design issues. Section 3 discusses the high level online. Except for the elimination of time and system modules. Lastly, this paper will be space barrier, this online content provides no concluded in section 4 and some future research enhancement to the learning experiences. work is suggested.

2. Architectural Considerations 2.1 Framework Design The knowledge environment will be This section presents the skeleton document developed using the layered architecture as shown that span the entire design process of the in Fig. 1. The layered architecture is chosen over environment. Although this research addresses other architectural patterns such as client/server, problems only in the domain of e-learning, the three-tier or peer-to-peer pattern as it aids in system must be modeled with an architecture that organizing the systems into layers of subsystem. is flexible enough to be independent on any This is extremely important as organizing the infrastructure. Considerations must be taken to design model in layers aids reuse, one of the most avoid the pitfalls of CASE where a thousand important aspects of object oriented design (and different modeling languages are conjured, e-learning). Besides reusability, portability is also causing the problem to be escalated. enhanced as changes are isolated only to one layer. To cater for e-learning, the system must be Each layer in the layered architecture is a set open in its architecture and supports a large and of subsystems that share the same degree of extensible class of distributed digital information generality and interface volatility. services. The basic entities to be stored, retrieved Application-specific and managed will be information stored in the layer form of learning objects. In the system’s context, these learning objects will be seen technically as a Application-general form of digital objects to be stored, accessed and layer managed. Thus, the basic function is to provide a Middleware access protocol service for locating and layer disseminating these digital objects. System-software This basic function will constitute a minimal layer set of requirements for the system to be fully Fig. 1 functional. While it is not the research’s intent to build a universal, wide-area digital information 2.2 Functional Layers infrastructure, considerable design efforts are The knowledge framework will exploit the employed to ensure that the infrastructure, at its capabilities of open-standard Web technologies very least, must be interoperable, robust, and such as XML and Web Services to construct a flexible enough to support other advanced Grid infrastructure that will harness the Semantic services and elaborations that will come into Web advantages. The Knowledge Framework existence as the system matures. Thus, the system encompasses three functional layers, namely, wire, will be built upon an interoperable layer that will description and discovery layer. not restraint the higher level user and service level choices such that it will become inappropriate to Discover y WSIL UDDI OWL+OWL Schema/DAML+OIL add in new features at any point of time. GSDL

With the primarily aim set at creating a Description WSDL RDF + RDF Schema platform for supporting social, intellectual, and technological , the knowledge system

will harness the capabilities of Grid Computing, SOAP / XML Protocol Semantic Web and Web Services. These three Wire Protocol technologies are seen as the future IT XML technologies that will spearhead the knowledge revolution. Applications will be developed to Fig. 2 illustrate the environment’s generality and efficiency in the sharing of resources. A 2.2.1 Wire Protocol Layer collaborative workspace for supporting diversified This layer forms the platform and specifies the virtual, Grid computing and e-learning wire protocol for supporting the client and server applications will also be provided. running on different in the distributed

network. Among the numerous wire protocols, Description Framework (RDF) and Resource Simple Object Application Protocol (SOAP) is Description Framework Schema (RDFS) to chosen as it is seen as the technology that will be achieve “ understandable” information. deeply embedded in the future of distributed Although RDF is used as the ‘descriptive computing. Besides, SOAP offers ease of language’ for resources, it is a further W3C interoperability and scalability when used with initiative, building upon earlier developments Web Services, one of the future IT revolutions. such as Dublin Core and the Platform for Internet Thus, with the ease of coupling with other Content Selectivity (PICS) content rating technologies such as the Universal Discovery, initiative. Thus, while RDF is used to empower Description, and Integration (UDDI) and Web creation, exchange and the use of metadata, the Services Description Language (WSDL), SOAP is framework will use the Dublin Core Element Set chosen as the wire protocol to transform the to capture a representation of essential aspects knowledge applications and aids communication related to the description of resources. over the Web with the concept of Web Services. This layer also includes XML Namespace However, as SOAP is merely a wire protocol, it (XML-NS), a construct which is adopted in RDF does not implement security. Thus, we are to provide a comprehensive system describing all proposing the use of SOAP with Extensible resources’ aspects. XML-NS provides a simple Markup Language (XML) protocol to allow the method for qualifying elements and attribute employment of application-level security. names by associating them with namespaces As SOAP is a text-based protocol that uses identified by URI references. It harnesses the XML for data encoding, XML will be employed great power of XML and RDF to declare their at the of the framework. XML is a cross- own modes of expression for resources platform, extensible, and text-based standard for description and yet allows 'authorship' sharing. representing data. It is also a key technology in Thus, the framework will employ Dublin Core the development of Web Services. Element Set when possible and use XML/RDF to The following standards are employed: XML allow terms in a metadata vocabulary to be treated Version 1.3, and SOAP Version 1.2. as reliably, unambiguous, describable objects. Besides providing description, this layer also 2.2.2 Description Layer looks into describing the communication of The Description Layer lies directly above the protocols and message formats. To provide a Wire Protocol Layer. With a sound and solid standardized way to structure communication, platform exemplified by XML and SOAP or WSDL is used. WSDL defines an XML grammar XML Protocol, this layer serves as the general for describing network services as a collection of purpose description language layer for identifying communication endpoints capable of exchanging the required information. With the proliferation of messages. It is extensible and allows description information, it is extremely difficult to find the of endpoints and their messages regardless of the exact information in the giant information dump. message formats or network. In short, for Web This problem is further amplified when dealing Services, although SOAP offers the basic with different machines in the distributed communication, it does not state how messages networks. Besides the problem of locating and must be exchanged to successfully interact with a understanding the information, the proliferation of service. Thus, WSDL is used as the specification communication protocols and message formats to describe all available services through a set of posed another problem. Thus, this layer is endpoints operating on the message. included to address these problems. Aiming to incorporate the dynamic service To enable efficient retrieval, the environment nature of Grid, the Open Grid Services needs to understand the information’s semantics. Architecture (OGSA) is included to bring about a This is done through the use of metadata. convergence of the Grid and Web Services Metadata has the potential of transforming communities. As OGSA adopts the general Web information from ‘machine readable’ to ‘machine Services approach and protocols to build its own understandable’. Thus, we employed Resource Grid infrastructure, we will use Grid Service

Description Language (GSDL) on top of WSDL. environment and discovered using distributed Grid services described in terms of GSDL will protocol such as WSIL. WSIL is also an XML- serve as an additional building block to function based document format that allows the inspection, as an enhanced that extends the discovery and aggregation of Web Services conventional Web Service functionalities into the description in a simple and extensible . Grid domain. These Grid services will be While similar in scope to UDDI, WSIL is a published and registered in the Discovery Layer complementary but different model to service through specialized Grid registry and can be discovery. Unlike UDDI, WSIL approaches invoked by using Internet protocol such as SOAP. service discovery in a more decentralized fashion The following standards will be employed: with the assumptions that the service requester is RDF Version 1.0, RDF Schema Specification 1.0, already familiar with the service provider. DCMES 1.1 and WSDL Version 2.0. This layer also integrates the Semantic Web vision of machine interpretability of content by 2.2.3 Discovery Layer incorporating semantic meanings. Web Ontology The first two layers discussed so far are Language (OWL) is intended to be used explicitly required for interoperable Grid and Web Services. to represent the meaning of terms in vocabularies They serve as the basic infrastructure to enable and their relationships. It is built on top and specific services and applications to leverage the beyond the basic semantics of RDF-S and will be current Grid and Web Services over the Internet. used to formally describe the terminology. Thus, The next layer, the Discovery Layer will deal OWL goes beyond XML, RDF, and RDF-S in its primarily with the provision, publication and ability to represent machine interpretable content. discovery of those existing services for The following standards will be employed: knowledge discovery. These services can be WS-Inspection 1.0, UDDI, OWL, DAML + OIL. classified into 2 domains, namely, Grid and e- learning domain, and used the Web and Semantic Services for publication and discovery. This layer 3. System Components will cover the interaction between machines that agree on the protocols and services described by SOAP / XML Protocol the lower layers to aggregate the supporting and VOs Grid E-learning OHS more specific services or applications. Applications Computing Applications Transformers

Colla A C O I The Universal Description, Discovery and A S S S S S S S n o u A n u e e e e e e e f G r t e r r r r r r r t g h bor e g o v v v v v v v r

r m

o

e e l i i i i i i i i W M Integration (UDDI) and the Web Services S o d c c c c c c c r n n e i a e e e e e e e e g

t c a n n e

t

c s s s s s s s y e n i b g t

u v

a

r e

Inspection Language (WSIL) will be used as the g i

t e y m

a e n

mechanism to discover the service. Service n d t

Federated Knowledge Repository Discovery is defined as any action that enables a Open Grid Services Semantic Web Architecture service requester to find and access the WSDL Architecture documents. This process can be as simple as Web Services Architecture accessing a file or URL containing the WSDL or Operating System Environment as complex as querying a UDDI registry and Hardware and Network Environment (Standards and Protocols) using the WSDL file(s) to select potential services. UDDI is a specification for a registry that can Fig. 3 be used by a service provider to publish WSDL documents. Clients search the registry looking for The full system component is depicted in Fig. services and then fetch the WSDL documents. 3. However, limited by the scope of this paper, Building upon the basic platform of XML, XML- only the e-learning modules will be discussed. In S and SOAP, UDDI provides a foundational the following section, we will discuss the key e- infrastructure for a Web Services based learning services, namely the Ontology Services, environment for the Grid and e-learning domain. Inference Services, Collaborative Authoring Alternatively, the service descriptions can be Services, Augment Services and Agent Services. published locally within the service hosting

3.1 Ontology Services Although there exists a number of publicly To serve as an interoperating medium for available commercial ontologies (e.g. UNSPSC, communication, explicit specification of learning RosettaNet) and educational ontologies (e.g. resources must be established. A common DMOZ, KSL), the system also acknowledges that understanding is vital for different machines to sometimes, no relevant ontologies for a particular use and share knowledge. This is established in subject exist. Thus, the SMEs can also develop a the form of ontology. Ontology is an explicit new ontology from scratch using Protégé. specification of some topic. It is a formal and Although, there is no single correct ontology- declarative representation which includes the design methodology, a rather comprehensive vocabulary for referring to the terms used in a guide exists in Ontology Development 101 [3]. particular subject area and the logical statements that describe what the terms are, how they can or 3.2 Inference Services cannot be related. In our context, ontology is The conceptual logics of the system will be defined as a formal explicit description and implemented using the fundamental theory behind representation of the learning resources and their Language Engineering (LE). LE is the discipline relationships in a domain of discourse. It will or act of engineering software systems that contain vocabulary that stores machine- perform tasks involving the processing of human interpretable definitions of the concepts and its language. Both the construction process and its relations among them. Employing ontology for outputs are measurable and predictable [4]. It is communicating knowledge will put the system in the application of Natural Language Processing line with the Semantic Web Vision whereby (NLP) to the construction of computer systems information is given explicit meaning, making it that process language for some task usually other easier for machines to automatically process and than the modeling language itself. LE will be used integrate information available on the future Web. to create knowledge on the language (partially With OWL employed as the ontology through the use of ontology) to enhance the query language, the system will use the open source and retrieval capabilities of the system. It will Protégé-2000 tool as the system’s ontology and enable the system to access the learning resources knowledge-base editor. Protégé is an open-source, efficiently and focus precisely on the exact Java tool that provides an extensible architecture information that the learner needs, thereby, saving for the creation of customized knowledge-based time and avoiding information overload. This will applications. Although it is currently only being also undoubtedly enhance the learning experience. used in clinical medicine and biomedical sciences, The inference services will enhance and aid it can be used in any field where concepts can be the query and search capabilities. For efficient modeled as a class hierarchy. As an ontology retrieval, understanding the query is obviously the editing tool, the system uses the Protégé's OWL most fundamental step. However, prefect Plug-in to describe the ontology declaratively, understanding of the whole query is not always state explicitly what the class hierarchy is and to necessary as the process is time consuming and which class it belong. complicated. Instead, our system employs a basic Employing a concept-based approach, search function that first requires the learner to concept will be described by class. Class will be limit the scope of query. With the scope limited, the main focus of the ontology. Thus, ontology the system then performs a shallow or partial together with individual instances of classes analysis of the query through keywords extraction. constitutes a knowledge base. However, as This partial understanding is often a useful ontology development and design is not the main preliminary step because it makes it possible to be research focus, this research will not focus on its intelligently selective before taking the depth of development and instead adopt standardized understanding to further levels which is usually libraries of Web ontologies. We will import time consuming. Thus, with an initial keywords ontologies from Ontolingua ontology library, list, the system will try to match each keyword DAML ontology library and/or OpenCyc Selected with the metadata repository (using ontology). Vocabulary and Upper Ontology. With the semantic nature and structure of the

learning resources coupled with the domain “Mechanics” and “Quantum Mechanics”, the identification, this initial analysis should be able system will also be able to associate semantically to generate results of high precision. close concepts such as “Wavelength-Momentum However, should the initial analysis fails, the Relation”, “Schrodinger’s Equation” or “Wave system could also perform a deeper semantic Functions” to Quantum Mechanics although these analysis using the Advanced Search. The concepts might not contain the keyword at all. Advanced Search uses the inference engine to refine and provide the semantics reasoning. Thus, 3.3 Collaborative Authoring Services the learner needs not process any prior knowledge This service is designed with the basic on the query topic nor its domain and is free to concepts of the ADDIE Model and modified to enter a string of unconstrained texts that he feels cater to the learner-centric environment in the e- best describes his problems. This raw query will learning context. ADDIE is an acronym referring be parsed using NLP techniques and the main to the major processes that comprises the generic keywords and rules will be passed into the Instructional Systems Design (ISD) process: inference engine for reasoning. Analysis, Design, Development, Implementation, The inference engine will then be responsible and Evaluation. It is one of the most popular for creating a conceptual domain of the query and instructional design models and has been provide possible reasoning over the conceptual considered by many e-learning professionals as representation. It then performs dynamic semantic the standard instructional model [7]. This model is analysis on the search criteria and matches them a common model used for developing training against the predefined semantics of the learning programs and has been documented, examined resources, thereby, giving a far better result than extensively and used successfully by many simple keyword searches. instructional teams over the years. The knowledge representation and reasoning This service makes use of a content services will be carried out using Description development model that is developed in an earlier Logic (DL). DL has been used in a range of work. This methodology which is based on the applications and reasoning with database schemas modified ADDIE Model, presents a systematic and queries [5]. They are also used widely as a approach to develop the concept’s contents using formal basis for ontology language and to provide the developed content development template. reasoning support for ontology design and deployment. To exploit the semantics of ontology Collaborative Authoring based relations, the DL-based inference engine must be able to reason with the ontology language, Services OWL, which is proposed in the earlier section. Therefore, the inference engine will perform Content Content reasoning with OWL through the use of OWL DL. Development Model Development Template Some key inference properties identified are Consistency Check, Subsumption, Equivalence E A Check, Instantiation Check and Retrieval. Analysis Guidelines Besides DL, Latent Semantic Indexing (LSI) Design for the 5 Development processes in will also be used to aid reasoning. LSI [6] is a Implementation I Evaluation D the ADDIE well-defined mathematical method, which uses Model pure mathematics to create the semantic cohesion D between documents and collections of documents. Fig. 4 Besides pure keyword matching, LSI also employs indexing to find semantically close 3.4 Augment Services concepts that are associated with keyword. For Knowledge augmentation will be fulfilled example, when a learner wishes to learn through the use of resource annotation. Initially, “Quantum Mechanics”, instead of searching the system will employ Amaya as the Web editor solely for the keyword of “Quantum”, to create rich annotation content for each concept.

3.5 Agent Services ii. He can browse the course materials and select Agent technologies have been used any concept that he wish to concentrate. extensively for planning and designing interactive iii. He can also view all the related course and collaborative communication [8, 9]. In this concepts or retrieve the additional course paper, we designed an agent service to provide the supplement map to get a broad understanding. means to create a personalized learning iv. Upon selection of a particular concept, he can experience. Personalization is catered for in two choose the required level of abstraction. stages. The first stage generates the exact learning The second scenario describes a learner who paths that the learner has to take to master a wishes to enhance his knowledge on a particular particular concept. These paths will be generated domain in which he has not enrolled in. dynamically based on the learner’s expertise (through mapping with his cognitive structure) i. Once the learner logs in, his preset and the requested concept’s expertise. Initially, preferences and constraints will be retrieved. each new learner will be given an empty academic ii. He will use the domain specific ontology to record categorizes by the subjects domain that is search for the concept that he wishes to master. offered by the system. Each domain is an iii. That concept will be assigned as the end point externalization of the learner’s cognitive structure. of his personalized learning path (EP). The map initially empty will be filled with each iv. Limited to the specified domain, the system successful acquisition of knowledge (through the will first search and retrieve the entire domain completion of post-assessment). Thus, an exact concept map that the requested concept resides. learning path can be mapped easily by comparing v. Next, the learner’s expertise in the specified the domain concept map with the learner’s domain will be retrieved. His prior knowledge cognitive concept map. The next stage is to select and expertise in this particular domain which is the type of learning approach to present the stored in his cognitive structure will be content. As a learning objective is realized with externalized and presented in a concept map different content approaches and assessment form. This cognitive map will then be used as a styles based on the Index of Learning Styles (ILS), comparison with the basic domain concept map the system will dynamically match the learner’s to find the next most immediate pre-requisite preferences to the ILS and select the type of concept that is present in both maps. learning experience that best suit the learner. vi. The identified concept will be assigned as the The following illustrates two learning starting point of the learning path (SP). scenarios that show how the personalized learning vii. The system will route dynamically and paths can be generated dynamically. The first provide all possible routes from SP to EP. scenario describes a learner who has enrolled in a viii. With the routes generated, the system will course and wishes to retrieve either the course then retrieved the learner’s preferences and pick materials or some additional reading materials to the content presentation approach (based on ILS) supplement the course material. In this scenario, that best suit the learner’s learning preferences. the lecturer has to specify a course map as well as ix. Lastly, depending on the system’s restrictions a supplement course map. The course map will (such as course fees or the lecturer’s cover all the key concepts that are pre-requisite to recommendation) and the recommended scope, the completion of the course, while the the allowed routes are presented to the learner. supplement course map will provide some additional reading materials. The system will The routes will be displayed in the form of a allow the learners to access all these materials meta-concept map. That is, upon selecting any without checking for any pre-requisite one of the nodes in the meta-concept map, the requirements as it is the lecturer’s duty to ensure selected node will act as a hyperlink and brings that these learners possess the prior knowledge. the learner to the beginning of either a course or a module. When the node is hyperlinked to a course, i. Once the learner logs in, his preset the course concept map will be presented. At this preferences, and constraints will be retrieved. stage, the learner can then begin his learning

process by selecting the course nodes which will [4] H. Cunningham. A Definition and Short bring the learner to his personalized module. On History of Language Engineering. Journal of the other hand, if the node of the meta-concept Natural Language Engineering, pages 1--16, vol map corresponds to a module, the learner can 5, 1999. immediately begin his personalized learning by [5] D. Calvanese, G. De Giacomo, and M. clicking on that node. Lenzerini. On the decidability of query containment under constraints. In Proc. of the 4. Conclusion and Further Work 17th ACM SIGACT SIGMOD SIGART Symp. on Principles of Database Systems PODS'98), In this paper, we described an approach to use pages 149-158, 1998. a concept-based knowledge environment to [6] Deerwester, S., Dumais, S. T., Landauer, T. address the lack of personalization, collaborations K., Furnas, G. W. and Harshman, R. A. (1990) and the failure to adopt technological "Indexing by latent semantic analysis." Journal of advancements to enhance learning. We have the Society for Information Science, 41(6), 391- shown how we can incorporate Grid Computing, 407 Semantic Web and Web Services into our [7] Betsy Bruce (2003) eLearning with knowledge framework. Embedding these Dreamweaver MX: Building Online Learning technologies into the e-learning pedagogy, we Applications. Macromedia Press. pp11-31. discuss how the system is able to provide for the [8] C.-Y. Chen, W.-S. Lo (2004). An Agent e- personalization of learning experiences as well as Learning System for Interactive and Collaborative the exploitation of the semantics for efficient Communication. In Proc. of the 3rd WSEAS Int. searches and retrieval of learning resources. Conf. on Artificial Intelligence, Knowledge Research efforts will be ongoing to revise the Engineering, Data Bases (AIKED 2004) system design to include new research methods to [9] S. Saberi, M. Kharrat, K. Badie (2004). Using improve the system. One key area will be Case-Based Planning in Designing and Planning a pedagogy research. As the driving motivator for Multiagent System. In Proc. of the 3rd WSEAS creating a successful enhancement in learning Int. Conf. on Artificial Intelligence, Knowledge experience should be based on pedagogy rather Engineering, Data Bases (AIKED 2004) than technology [10], pedagogy research effort [10] Dexter, S. L., Anderson, R. E. & Becker, H. J will be centered on analyzing new methods to (1999). Teachers’ views of computers as catalysts enhance the learning experiences of the learners. for changes in their teaching practice. Journal of Research on Computing in , 31 (3), References: 221-239.

[1] Miltiadis D. Lytras, Nancy Pouloudi (2001). ‘E-learning: Just a waste of time’. Association for Information Systems, 2001 In Strong, D., Straub, D. & DeGross, J.I. (Eds) Proceedings of the Seventh Americas Conference on Information Systems, (AMCIS 2001) [CD Proceedings], August 3-5, 2000, Boston, Massachusetts, USA pp. 216-222. [2] Pithamber R. Polsani: Use and Abuse of Reusable Learning Objects. In JoDI - Journal of Digital Information, volume 3 issue 4. Themes: Information management. 2003-02-19. [3] N. Noy and D. L. McGuinness. Ontology Development 101: A Guide to Creating Your First Ontology. Knowledge Systems Laboratory, March, 2001.