Eighth IEEE International Conference on Advanced Learning Technologies Towards a Knowledge Portal for E-Learning based on Semantic Web* Yanyan Li1, Mingkai Dong2 1 Knowledge Science & Engineering Institute, School of Education Technology, Beijing Normal University, 100875, Beijing, China 2 Knowledge Management, Siemens Corporate Technology China Liyy1114@gmail.com, mingkai.dong@siemens.com Web encompasses efforts to build a new WWW architecture that supports content with formal semantics, enabling better searching and navigating through the cyberspace. Thus, the Semantic Web can be used as a technology for realizing sophisticated elearning scenarios [5]. Based on Semantic Web technologies, this paper proposes a knowledge portal to effectively support elearning, enabling flexible knowledge acquisition, knowledge refinement and maintenance, as well as knowledge retrieval and accessing. Abstract E-Learning aims to deliver individualized, comprehensive, dynamic learning content in real time, aiding the development of communities of knowledge. By incorporating Semantic Web technologies, this paper proposes a knowledge portal with underlying distributed learning repositories to effectively support the knowledge process, empowering flexible knowledge acquiring and collecting, knowledge refinement and maintenance, as well as thematic knowledge retrieval. In this way, the knowledge portal serves as an entry point for learners to access and utilize the semantic related knowledge to meet their diverse learning needs. 2. Architecture The architecture of the knowledge portal for elearning is illustrated in figure 1. As the figure shows, the instructors and learners can access and maintain the learning resources through the user interface. Reference ontology indicates the common terminology with respect to the specific domains, which lays the foundation for semantically interconnecting the distributed learning repositories. Other core modules includes: knowledge acquiring, knowledge refinement, and knowledge retrieval. 1. Introduction With the popularity of the e-Learning, Learning Content Management System (LCMS) emerged expecting to provide standard-based content repositories that allow learners to capture, store, deliver, and manage learning resources. But the substantial impediment to the destination is the fact that the resources are disordered, isolated, and heterogeneous, and there is no common overarching context for the available resources. Additionally, finding the precise information is very difficult because of the lack of semantic description of learning resources. Thus, navigation through a large set of independent resources often leads to users’ being lost. The literature in elearning research emphasizes the value adding learning from the knowledge management perspective [2] [3]. Therefore, how to acquire and provide the suitable knowledge to meet learners’ diverse learning needs is a key issue. Knowledge portals are portals that focus specifically on the production, acquisition, transmission or the management of knowledge, which emphasizes the consideration of the relevant context of information providing and accessing [4]. The Semantic * Figure 1. Architecture of e-learning knowledge portal. This research work is supported by National Natural Science Foundation of China (NSFC: 60705023) 978-0-7695-3167-0/08 $25.00 © 2008 IEEE DOI 10.1109/ICALT.2008.83 910 reasoning rule: R(K1, K2), R(K2, K3), …, R(Kn-1, Kn), ⇒ R(K1, Kn), where R ∈ Ω, Ω is a set of primitive semantic relationships in a given domain [7]. 3. Semantic Resource Organization Model The model is composed of knowledge algebra and knowledge base. The knowledge algebra defines a set of basic elements (vocabularies), which provides a foundation for knowledge base. The knowledge base comprises two parts: knowledge schema and fact base. The knowledge schema describes the agreed abstract domain knowledge (e.g. axioms, rules), while the fact base describes the facts of instances (e.g. types, properties). Actually, the knowledge base is modeled as a directed labeled graph, wherein the nodes and arcs respectively represent the entities (e.g. concepts, instances) and semantic relationships between entities, encoded as RDF entities and properties. An example works best to explain our proposed idea. Suppose we are trying to establish one semantic-based model in “research” domain, which roughly explains how we can organize resources for research purpose. Figure 2 is an excerpt from a simplified resource model. As the figure illustrates, the knowledge schema is composed of abstract concepts with various semantic relationships while each concept has its specific properties. The instances can be dynamically linked to the concepts in the knowledge schema. 4. Knowledge Process for E-Learning Adoption 4.1 Knowledge Acquiring Mechanism Two modes are provided to acquire the knowledge in a certain context. Through the authoring interface, instructors and experts are able to define the generic instructional strategies and metadata (e.g. content, structure, and context) of learning materials based on templates. Through this annotation process, users can create metadata that conform to the domain ontology. On the other hand, interaction in an e-learning environment implies a lot of knowledge creation and exchange. Thus, data mining technologies are adopted for recognizing and tracking the topics in the interaction process. Discussion forums allow learners to ask and answer questions on a variety of topics, which is especially useful for discussion involving multiple viewpoints or multiple problem solutions, or for straightforward questions requiring a quick response. RDF and XML are used to represent and save the acquired knowledge and it’s linking resources. 4.2 Knowledge Refinement Knowledge refinement comprises relevance checking and redundancy elimination. The relevance checking module determines whether the knowledge items (e.g. questions, metadata) link to the appropriate topics in domain ontology. We adopt the SVM [6] to classify the texts into appropriate classes. Simultaneously, the Knowledge Engineer checks the relevance of knowledge so as to guarantee the correct classification of knowledge items. There are two types of knowledge redundancy: explicit repetition, and implicit redundancy, i.e., knowledge can be derived from other existing knowledge. As some implicit redundant knowledge may enhance the efficiency of a system, we currently only deal with the explicit case. As for the questions in discussion forums, we design many question templates and group into clusters according to their meaning. A thesaurus consisting of noun-abbreviation, composite word, and verb-synonym, etc. is also used. For example, define, describe and introduce concern the explanation or definition of concepts. We currently consider two factors (i.e. keyword focus and sentence structure) in determining whether there have redundant questions or not. If redundant questions exist then mark the appearance times whilst deleting the duplicates. The Knowledge Engineer will assist checking the knowledge redundancy and consistency. Figure 2. A schematic resource organization model. Especially, the instances belonging to the same concept may be correlated with typical semantic relationships, such as Sequential, SubtypeOf, CauseOf, SimilarTo, Reference, etc. More types of semantic relationships can be defined according to the application domains. The reasoning rules can be used for chaining the semantic relationships and obtaining the reasoning result from the chaining. A simple case of the reasoning is that all the semantic relationships have the same type, which is called single-type reasoning. According to the transitive characteristic of the semantic relationships, we have the following 911 the paths. By utilizing domain knowledge and context, the algorithm Get_AssociationPath is to discover the possible semantic association paths between two entities. As a result, the semantic related knowledge items are structured and return to learners, providing an overall view on a topic along with the navigation guide for learners to explore more related knowledge. 4.3 Thematic Knowledge Retrieval Thematic knowledge retrieval enables learner to express his information needs in terms of keywords, but at the same time uses the semantic information regarding the domain of the application to obtain results that are not possible in traditional searches. It mainly comprises the following two steps. Get_AssociationPath(entity a, entity b, Integer h) { //h denotes the pre-defined constraint of maximum searching depth, A denotes the set of semantic association paths, L denotes the search depth. Qa = {a}; Qb ={b}; // Qa and Qb denote the searching sequence of a and b, respectively Va={a}; Vb={b}; // Va and Vb denote the set of the visited entities, respectively A=∅;L = 1; While (L<h) { While (Qa ≠ ∅){ // Pop all the elements in Qa to the temp sequence Ta tmp = pop(Qa); Push(Ta, tmp); } While (Ta ≠ ∅){ //repeat the following steps till Ta is null k = pop(Ta); if (k Va) break; //skip to the next node if k has been visited For i=1 to m { //m denotes the number of properties of k, pi denotes the ith property of k Vb) then { // vi denotes the value of the If (pi(k, vi)∩vi property pi If (vi = b) then{ There is an association path originated from a to b and add it into A. }Else { There is a joint association path between a and b and add it into A. } } If (pi(k, vi)∩(pi is ObjectProperty)) Qa= Qa + vi; Va = Va + v i ; } } The same operation on Qb; L = L + 1; } // end of “While (L<h)” Return A; } //end of algorithm 5. Conclusion The kernel idea of this paper is to semantic modeling of heterogeneous learning resources in a coherent and meaningful scheme. With the effective support of knowledge acquiring, knowledge refinement as well as knowledge retrieval mechanism, the proposed knowledge portal aims to provide an integrated, scalable and easy-to-use interface for diverse learners to manage and exploit the dynamically growing learning resources on the Web. The proposed knowledge portal is still in its early stage and more work needs to be done in the future, such as (1) Use semantic link network (SLN) to organize the semantic-associated learning resources supported with semantic mapping; (2) Investigate the topic recognition and tracking by considering semantic context; (3) Implement the knowledge portal prototype and apply in practical settings. ∈ ∈ z z 6. References [1] [2] [3] Query annotation. According to the learner profile, the inputted query is firstly annotated to denote the intended meaning of the learners by dealing with the linguistics ambiguity [1]. Select target knowledge items. Taken the matching entities as the anchor ones, this step is to find semantic relevant entities. The simple approach for selecting the target entities for the one matching entity, purely based on the structure of the graph, is to collect the first N triples originated from the anchor entity, where N is the pre-defined traversal constraints. As for the case of two matching entities corresponding to the query, it is the key problem to find all the semantic association paths between the two entities so as to select the relevant instances on [4] [5] [6] [7] 912 Li, Y., Huang, R., Semantic-based Thematic Search for Personalized E-Learning, LNCS4018, pp. 354-357, AH06, Ireland, 2006. Lytras, M.D., Doukidis, G.L., Skagkos, T.N., Elearning Pedagogy: A Value Definition from a Knowledge Management Perspective, Proceedings of the 5th annual TechEd International Conference, March, South California, USA, 2001. Maurer, H. & Sapper, M. (2001). E-Learning Has to be Seen as Part of General Knowledge Management, World Conference on Educational Multimedia, Hypermedia & Telecomuunications, Tampere, AACE, Charlottesville, VA, pp. 1249-1253. Staab, S., Knowledge Portals. KI 16(1): 38-39 (2002) Stojanovic, L., Staab, S., Studer, R., E-Learning based on the Semantic Web, World Conference on the WWW and the Internet, Orlando, Florida, USA, 2001. Vapnik, V., Golowich, S., Smola, A., Support Vector Method for Function Approximation, Regression Estimation, and Signal Processing, Advances in Neural Information Processing Systems, Cambridge, MA: MIT Press, 281-287, 1996. Zhuge H., Li, Y., Learning with active e-course in knowledge grid environment, Concurrency and Computation: Practice and Experience, 18(3), pp. 333356, 2006.