An intelligent semantic e-learning framework using context-aware Semantic Web technologies

Recent developments of e-learning specifications such as Learning Object Metadata (LOM), Sharable Content Object Reference Model (SCORM), Learning Design and other pedagogy research in semantic e-learning have shown a trend of applying innovative computational techniques, especially Semantic Web technologies, to promote existing content-focused learning services to semantic-aware and personalised learning services. To facilitate this transforming process, this paper presents a novel context-aware semantic elearning approach to integrate content provision, learning process and learner personality in an integrated semantic e-learning framework. As the basis of the computational framework, a scalable and extensible generic context model is proposed to structure the semantics of contextual relations and concepts in various contexts, such as learning content description, learning model, knowledge object representation and learner personality. Corresponding technical and pedagogical developments of this framework also consider compatibility issues with existing technologies (eg, XML/Resource Description Framework) and specifications (eg, IEEE LOM) in order to achieve the best interoperability. Introduction Internet-based e-learning revolutionarily changed the training and education industry for its uninterrupted online service for 24/7 access anywhere. Since the concept was formally introduced at the ‘Internet-based training’ workshop at the American Society for Training and Development 1996 international conference, the e-learning industry has gone through a boom-and-bust cycle of development (Kruse, 2002). Recent critical 352 British Journal of Educational Technology Vol 37 No 3 2006 © 2006 The Authors. Journal compilation © 2006 British Educational Communications and Technology Agency. analysis revealed the reasons behind the scenes from various perspectives. Woodill (2004) and Bunis (2003) point out that in addition to business and marketing reasons, there are several other important issues related to e-learning solution developments, such as the lack of learner-centric usability and interactive involvement and the lack of understanding of multimedia-aided learning and teaching. There is a demand for shifting e-learning solutions from pure web-based content provision to instructional and learner-centric learning and teaching environments. With the fast development of XML-based technologies on the Internet, the next generation of the World Wide Web (WWW)—the Semantic Web (Berners-Lee, Hendler & Lassila, 2001)—is starting to shape up. Ontology-based technologies and intelligent agents are expected to assist semantic information processing on the future Semantic Web. With more semantic-aware computing technologies, e-learning is expected to be more intelligent in the new era of Educational Semantic Web (Anderson & Whitelock, 2004). To facilitate the transforming process towards the Educational Semantic Web vision, this paper presents a novel context-aware semantic e-learning approach, to integrate content provision, learning process, and learner personality in an integrated semantic e-learning framework. The proposed e-learning framework supports intelligent semantic e-learning by (1) bringing semantic context awareness into multimedia learning information processing and learning practices, and (2) bringing awareness of learner personality in support of personalised learning. As the computational basis of the semantic e-learning framework, a scalable and extensible generic context model is proposed to structure semantics of contextual relations and concepts in various contexts such as learning content description, learning model, knowledge object representation, and learner personality traits. This context model aims to lower the knowledge barrier of semantic annotation of learning resources, which improves the usability of semantic e-Leaning systems. To achieve a high level of interoperability, compatibility of this context model with other existing technologies (eg, XML/ Resource Description Framework [RDF] [http://www.w3.org/RDF]) and specifications (eg, IEEE Learning Object Metadata [LOM] [http://ieeeltsc.org/wg12LOM/]) is considered in implementation. Hence common instructors and learners are not required to understand complicated concepts of ontology and reasoning with the existing Semantic Web technologies as a prerequisite before using knowledge-oriented services. Although this paper mainly focuses on providing a computational semantic e-learning solution, both technical and pedagogical issues related to the semantic e-learning are considered in the development process. Related work Learning content description In recent years, there has been a lot of effort put into learning content description standardisation. Among these specifications, IEEE LOM is the most popular one Context-aware semantic e-learning framework 353 © 2006 The Authors. Journal compilation © 2006 British Educational Communications and Technology Agency. adopted in most learning management systems (LMS). LOM aims at enabling learners or instructors to search, evaluate, acquire, share, exchange and utilise Learning Objects across technology platforms and systems. It specifies a set of metadata elements to guide the description and operation of learning objects at the conceptual level. However, from a user’s point of view, with such a big set of 47 elements in 9 categories, it is not easy to use the complete set, or just part of it, correctly in real content annotation practices. There are a number of dedicated LOM editors in research and development, which include the LOM Java Editor from Darmstadt University of Technology, Germany (http://www.multibook.de/lom/), the TreeLOM from Cukurova University in Adana, Turkey (Cebeci & Erdo~ an, 2005), the ImseVimse from the Royal Institute of Technology (http://kmr.nada.kth.se/ imsevimse/), Sweden and the LOM Metadata Editor embedded in Authorware from Macromedia. In the current e-learning industry, most LMSs work in a closed-system manner. Some systems still use their own framework for learning content description rather than adopting LOM as the main standard. These minority frameworks include TArgeted Reuse and GEneration of TEAching Materials (TargeTeam) (http:// www.targeteam.net/), Tutorial Markup Language (TML) (http://www.ilrt.bris.ac.uk/ netquest/about/lang/) and Procedural Markup Language (PML) for multimedia presentations (Ram, Catrambone, Guzdial, Kehoe, McCrickard & Stasko 1999). Given such a situation, even the same learning content (eg, an open access resource on the Web) could end up with a number of incompatible descriptions in different LMSs. Without a properly designed semantic interoperation interface, resolving the heterogeneity problem will take extra effort in practice. In addition to plain learning metadata, more complex semantics of multimedia resources are to be represented and managed in modern e-learning solutions. In the field of semantic description, the most important work is known to be the RDF—the basic information encoding language of the new Semantic Web. Other high-level languages such as DARPA Agent Markup Language (DAML) (http://www.daml.org/) are developed based on RDF, and most ontologies on the Web are now encoded in RDF as well. As a matter of fact, most recent research into applying Semantic Web technologies in e-learning are involved with RDF and ontology (Sampson, Lytras, Wagner & Diaz, 2004). However, RDF only provides a fundamental language for semantic description, and further developments of semantics capture and management are to be carried out beyond RDF. Sheth, Ramakrishnan and Thomas (2005) categorised three kinds of semantics to be captured for the Semantic Web: the implicit semantics, the formal semantics and the powerful (soft) semantics. However, working with semantics and knowledge needs awareness of context in application, for example, user context and working context (Hadrich & Priebe, 2005). To bring Semantic Web technologies into elearning, a lot of interesting work has been done recently. For example, Nilsson (2001) explores the potential impacts of Semantic Web on e-learning; Henze, Dolog and Nejdl (2004) present a logic-based approach for resource representation and reasoning based on RDF annotations; Simic, Gasevic and Devedzic (2004) discuss how Semantic Web 354 British Journal of Educational Technology Vol 37 No 3 2006 © 2006 The Authors. Journal compilation © 2006 British Educational Communications and Technology Agency. technologies, especially ontologies, could be used to present semantics of course content and student information. In addition to semantic description, another important issue in semantics management is semantic integration. Heterogeneity is one of the characteristics of the information on the open Web, and this will by no means remain on the Semantic Web in the future. A similar situation exists in e-learning. Even if all learning content descriptions on the global network are in encoded in RDF, it still needs various levels of ‘bridge’ for semantic translation and interoperation, because from a pure computing point of view, semantic integration is an inevitable issue in distributed computing environments (Doan, Noy & Halevy, 2004). Because ontology is the most frequently used technique in semantic description, semantic services will have a challenge to face when the semantic integration issue emerges with multiple ontologies and schemas. Potential solutions from generic computing perspectives are normally adopted in specific application areas such as e-learning and e-business. Recent typical approaches include ontology mapping and integration (Akahani, Hiramatsu & Satoh, 2003; Silva & Rocha, 2003) schema manipulation (Bernstein, 2003; Embley, Xu & Ding, 2004) and semantic interpretation (Huang & Hacid, 2003). In the e-learning field, Gasevic and Hatala (2005) present an algorithm for ontology mapping in course description context. Pedagogy and learning process in e-learning In addition to learning content description and semantics management, another important field in semantic e-learning is the learning process and pedagogy support. Wikipedia (http://en.wikipedia.org/) defines pedagogy as ‘the art or science of teaching’. There are many aspects covered in the concept such as instructional design and theory, learning theory, and other social–cultural, ethical elements. In the education field, pedagogy is a traditional research area that evolves all the time, but pedagogy research in elearning context is a relatively new sub-area. In recent studies on pedagogy in e-learning, Allert (2004) gives out a comprehensive survey of metadata models used to present learning concepts in contexts from the social perspectives. Sicilia and Lytras (2005) investigate ontological structures for generic constructivist and sociocultural learning. Lakkala, Lallimo and Hakkarainen (2005) study the issues of pedagogical design in the context of a collaborative learning environment, including the effects of utilising new technology in web-based learning and the application of an appropriate method in the design process. Azouaou and Desmoulins (2005) present an ontology-based conceptual model for pedagogy and document annotation for teachers. Lama, Sanchez, Amarim and Vila (2005) present an ontology for IMS Learning Design (LD) concepts and learning activities. In terms of standardisation, LOM from IEEE and Sharable Content Object Reference Model (SCORM) from the Advanced Distributed Learning (ADL) Initiative (http:// www.adlnet.org/), the most popular e-learning standards, have not taken pedagogy support as one of their core issues in specification. As a run-time infrastructure for e-learning, SCORM now only enables learning content packaging, organisation and Context-aware semantic e-learning framework 355 © 2006 The Authors. Journal compilation © 2006 British Educational Communications and Technology Agency. delivery. However, a new specification is under development as LD from IMS (http:// www.imsglobal.org/learningdesign/), based on a pedagogy-friendly Educational Modelling Language (EML) from the Open University of the Netherlands (http:// eml.ou.nl/). LD addresses pedagogy issues in processes within ‘units of learning’ or whole tasks (such as a course). It also provides a pedagogical metamodel to support various didactical learning approaches (both objectivist and constructivist) (Hummel, Manderveld, Tattersall & Koper, 2004). Other than the LD specification, there is another research project that addresses the learning process issue—PALO from The Universidad Nacional de Educación a Distancia (UNED), Spain (http://sensei.lsi.uned.es/palo/). Whereas EML uses a metamodel approach to explicitly describe the pedagogical approach used, with PALO the pedagogy is implicit in the particular PALO template used. However, no matter which pedagogy framework is adopted in a LMS, it is vital to seamlessly integrate the pedagogy supports of learning processes and learning theories with learning semantics management and other components in the system. Learner personality and personalised learning service Personalised learning and teaching could be regarded as an ultimate level of instruction. Confucius, a great thinker, philosopher and educationist of China, presented a philosophical statement about 3 000 years ago. His philosophy in teaching is known as: ‘teach students in accordance with their aptitude, adjust measures to local conditions’ (Confucius, 1997). Recent studies in modern psychology also show similar results. For example, Heinström (2000) proves that learner personality influences learning strategies and learning outcomes in real practice. A study of student characteristics and computer-mediated communication from Wilson (2000) reveals that personality may influence academic success in unanticipated ways. Therefore, to achieve the best performance in learning and teaching, especially in self-directed or instructorled e-learning, it is essential to be aware of the learner’s aptitude and personality in context. According to psychological studies on personality, there are five basic dimensions of personality traits that are stable across the lifespan and directly related to human behaviour (Revelle & Loftus, 1992). These dimensions are extraversion, neuroticism, agreeableness, conscientiousness and openness to experience—and are also known as the Five Factor Model of personality (Carver & Scheier, 2004). Personality traits are expressed in learning styles, which are in turn reflected in learning strategies, which eventually produce a certain learning outcome (De Raad & Schouwenburg, 1996). Personality traits serve as directors or blocks for motivation and learning strategies (Blickle, 1996). To enable successful personalised e-learning, support from a learning service provision is also essential. From this point of view, personalised learning is understood as an adaptive learning service via learning portals in many personalised e-learning solutions (Brusilovsky & Nijhawan, 2002; Dolog, Henze, Nejdl & Sintek, 2004). The aim has been towards ensuring that learning content and process are tailored to meet the needs of individuals. In this process, user modelling and personal profiling are commonly used 356 British Journal of Educational Technology Vol 37 No 3 2006 © 2006 The Authors. Journal compilation © 2006 British Educational Communications and Technology Agency. to enable personalisation. Munoz, Palazzo and Oliveira (2004) try to use domain and content knowledge ontologies and student models to improve personalisation. Keenoy et al (2004) provide personalisation service in e-learning via adaptive user personal profiles, which include a history of recent user activities in learning. Simon, Mikl’os, Nejdl, Sintek and Salvachua (2003) provide a mediation infrastructure for learning services, where its personalisation service is also delivered via dynamic learner profiling using ‘personal learning assistants’. In addition to user profiling, Chen, Lee and Chen (2005) address the learner ability aspect in web-based learning in addition to traditional aspects such as learner preferences, interests and browsing behaviours. Intelligent semantic e-learning framework In this paper, an intelligent semantic e-learning framework is presented to address semantic information processing, learning process support and personalised learning support issues in an integrated environment. The contrast between semantic e-learning and traditional e-learning information flow is illustrated in Figure 1. Traditional web-based e-learning systems use a web browser as the interface. Through run-time learning environments (either compatible or incompatible with SCORM), users could access the learning objects, which are directly linked to multimedia learning resources such as lecture video/audio, presentation slides and reference documents. In the proposed semantic e-learning framework, in addition to the existing learning information flow, three new components are introduced to bring in more intelligence in elearning. These components are a semantic context model, intelligent personal agents and conceptual learning theories. By using intelligent personal agents, the framework could perform adequate personal trait information profiling and deliver personalised learning services according to the individual’s personality and interests. By applying a new semantic context model, semantic information for static resource and dynamic process description could be more easily encoded and retrieved across the current WWW and the future Semantic Web, referring to ontologies or knowledge bases if necessary. The context model also enables Figure 1: Architecture of a semantic e-learning framework in context Run-time learning environments Learning objects Context model Users