整合语义网技术到企业网络学习解决方案 Incorporating Semantic Web Technologies to Enterprise E-Learning Solutions

2008-08-10 00:00:00 评论(0)
这是一篇关于如何把语义网技术应用到企业内部网络学习解决方案的文章,完稿于2008年8月28日。
请分享到:


Incorporating Semantic Web Technologies

to Enterprise E-Learning Solutions

 

Wenchao He

School of Information System and Technology

Faculty of Informatics, University of Wollongong, Australia


  

1       Introduction

 

The Semantic Web is the emerging landscape of new web technologies aiming at web-based information and services that would be understandable and reusable by both humans and machines. Its advent and the relevant technologies, tools and applications provide e-learning development and research with new contexts and opportunities. Enterprise e-learning solution can particularly benefit from the incorporation of Semantic Web technologies because of the nature of enterprise context—it requires that not only learning per se but also business processes should be taken into account, and Semantic Web technologies can optimise the solution so that the business goals can be more easily and likely achieved. This paper will explain the benefits and the approaches of incorporating Semantic Web technologies to enterprise e-learning solutions, and use the Ernst and Young example to demonstrate how to propose Semantic Web technologies to a given e-learning solution.

 

2       Why Incorporate?

 

2.1     Enterprise Training’s Goals

 

The most important concern about enterprise e-learning is its economic benefits. Converting traditional training delivery methods to e-learning will save the firms a large amount of budget. IBM saved US $200 million in 1999, providing five times the learning at one-third the cost of their previous methods. Using a blend of Web-based (80 percent) and classroom (20 percent) instruction, Ernst & Young reduced training costs by 35 percent while improving consistency and scalability. Rockwell Collins reduced training expenditures by 40 percent with only a 25 percent conversion rate to Web-based training. Many other success stories exist (Strother 2002). Such economic benefits are usually measured by comparing e-learning and traditional learning, focusing on the cost of travel, training locations, learning time, etc (Hall 2002). However, the nature of the training, disregarding its delivery mode, has tended to be ignored. That is to increase the effectiveness and efficiency of employees’ work so that the firms’ management and business goals can be achieved.

 

When turning the focus back to the nature of training, we may want to rethink how e-learning can improve the employees’ performance, because this is the goal of “training”. Typical enterprise e-learning is predesigned and developed and the employees are to use and learn the self-paced instructional materials or attend synchronous sessions in which there is an instructor leads the “class”. A better scenario may be inviting the end-users to participate in the designing process so that the course may be more suitable for them (Carr-Chellman & Savoy 2004). However, if the learning content is not targeting the area that the learners are still not able to achieve but may achieve with assistance, the learning will be ineffective. That area is called Zone of Proximal Development (ZPD), which is the gap between what has been actually developed and what could be developed potentially (Vygotsky 1978). Different from typical educational settings where the teachers have time to observe and identify students’ ZPDs, enterprise e-learning designers and trainers tend to have much less time and effort on analysing exact learners. The mismatch between exact learners’ ZPDs and the e-learning content will lead to the low effectiveness, efficiency and motivation. Even though the learners have participated in design process where their prior knowledge are evaluated and then the course has been designed “appropriately”, they will still feel that either the e-learning content is repeating what they have known or teaching them irrelevant stuff as the enterprise contexts are changing from time to time. For example, the off-line communication between employees (e.g., informal chat, formal meetings, etc.) can increase their shared knowledge, but the predesigned e-learning system does not know about this. Then the employees are still required to go through specific e-learning units as it is part of the HR management procedures. In this case, the e-learning is wasting the staff’s time, i.e. the companies’ money.

 

To achieve the best training results, First Principles of Instruction suggest that (1) the learners should be engaged in solving real-world problems, (2) the existing knowledge should be activated as a foundation for new knowledge, (3) the new knowledge should be demonstrated to the learners, (4) the new knowledge should be applied by the learners, and (5) the new knowledge should be integrated into the learner’s world (Merrill 2002). Therefore, one of enterprise e-learning’s goals is that the content should match the individual learner’s context (e.g., the real-world problems he or she is facing, the prior knowledge, the possibility that he or she can really apply the new knowledge, etc). However, with traditional Web technologies, it is difficult for enterprise e-learning solutions to achieve that goal, unless there are trainers who can adjust the e-learning content and instructional strategies from time to time according to relevant information from learner analysis or learning evaluation, like the teachers in educational settings.

 

2.2     Semantic Web Technologies for E-Learning

 

Content-based and user-based techniques are two traditional approaches to personalization (Dai & Mobasher 2004). Recommendations produced with the content-based technique based on content similarity to the personal profile of the users, while the user-based techniques focus on similarities to other users (Mobasher, Jin & Zhou 2003). Their drawback concerns the difficulty to capture semantic knowledge of the application domain i.e. concepts and their relationships, inherent properties associated with the concepts, axioms or other rules, etc (Markellou, Mousourouli, Spiros & Tsakalidis 2005). As the Semantic Web comes with new emerging standards based on evolving Web technologies, it allows the reuse of material in different contexts, flexible solutions, as well as robust and scalable handling. For achieving this, the web documents are now annotated with meta-information or metadata. This metadata defines what the documents are about in a machine processable way. Ontologies offer a way to cope with these hererogeneous representations of Web resources. They comprise the backbone of the Semantic Web and appear as a promising technology for implementing in particular e-Learning applications. The reason ontologies are becoming so popular is due to what they promise: a share and common understanding of a domain that can be communicated between people and application systems. An ontology can formulate a representation of the learning domain by specifying all of its concepts, the possible relations between them and other properties, conditions or regulations of the domain. The development of the ontology is akin to the definition of a set of data and their structure. In this way, the ontology can be considered as a knowledge base that is used further for extracting useful knowledge and producing personalized views of the e-learning system, so the learning experience can be likely to comply with the First Principles of Instruction which are in direct relation to enterprise training’s goals. (Davies, Fensel, Harmelen & NetLibrary 2003).

 

3       How to Incorporate?

 

3.1     Personal Agents

 

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. To enable personalised learning experience with Semantic Web technologies, the first step is to construct personal agents (see Figure 1). Agents are pieces of software that work autonomously and proactively. Conceptually they evolved out of the concepts of object-oriented programming and component-based software development. A personal agent on the Semantic Web will receive some tasks and preferences from the person, seek information from the learning resources, communicate with other agents, compare information about user requirements and preferences, select certain choices, and give “answers” to the user. By using intelligent personal agents, the framework can perform adequate personal trait information profiling and deliver personalised learning services according to the individual’s job description, training history/record, personality, interests, etc (Antoniou & Harmelen 2004; Huang, Webster, Wood & Ishaya 2006).

 

Here are some examples of how the personal agents work:

 

(1)         Within the same department of a company, some employees are new staff while others have been working there for a long time. Their personal agents can determine to delivery them different learning content. The new staff may be required to learn more content, while the old staff may only need to update their knowledge according to any change of company policies, business process, etc.

(2)         Some learners prefer watching and listening while some others would like to read. The personal agents can record the preferences and then check the learning content’s availability in terms of forms/formats and then provide the learners with choices.

(3)         When learner starts a new learning journey, his or her personal agent can record all the activities and the performance (e.g., what has been understood and what needs to be enhanced and practised more in the future). If a new knowledge requires prior knowledge to support understanding, the personal agent may firstly check whether the learner meets the requirement and then determine whether give him or her access to the content.

 

Figure 1: E-learning System with Intelligent Personal Agent

 

3.2     Content Reconstruction

 

The second step is to reconstruct all the content by adding metadata, ontologies and logic. Metadata is machine understandable information for the web. It provides a common set of tags that can be applied to any resource, regardless of who created it, what tools they used, or where it’s stored. Tags are, in essence, data describing data. Metadata tagging enables organizations to describe, index, and search their resources and this is essential for reusing them. In the e-learning community three metadata standards are emerging to describe e-learning resources: IEEE LOM, ARIADNE and IMS. Those meta-models define how learning materials can be described in an interoperable way. All the metadata elements necessary to describe a resource can be classified into several categories, each offering a distinct view on a resource. For example, the LOM standard contains the following metadata levels:

 

(1)         General: groups all context-independent features plus the semantic descriptors for the resource;

(2)         Lifecycle: groups the features linked to the lifecycle of the resource;

(3)         Meta-metadata: groups the data elements describing the metadata that indexes the document;

(4)         Technical: groups data elements describing the technical features of the document;

(5)         Educational: groups educational and pedagogic data elements for the resource;

(6)         Rights: groups data elements pertaining to the conditions of use for the resource;

(7)         Relation: groups data elements that describe the linkage between the subject and other resources;

(8)         Annotation: groups data elements that allow comments on the educational use of the resources;

(9)         Classification: groups data elements that describe the position of the resource in an existing classification system.

 

All the e-learning content needs to be tagged with metadata so that the machines can understand the type of the content which may be matched by personal agents.

 

Ontologies are specifications of the conceptualization and corresponding vocabulary used to describe a domain (Gruber 1993). They are well-suited for describing heterogeneous, distributed and semistructured information sources that can be found on the Web. By defining shared and common domain theories, ontologies help both people and machines to communicate concisely, supporting the exchange of semantics and not only syntax. It is therefore important that any semantic for the Web is based on an explicitly specified ontology. By this way, learners’ and instructional designers’ agents can reach a shared understanding by exchanging ontologies that provide the vocabulary needed for discussion. At present, the most important ontology languages for the Web are the following (Antoniou & Harmelen 2004):

 

(1)         XML

 

(2)         XML Schema

 

(3)         RDF

 

(4)         RDF Schema

 

(5)         OWL

 

Logic is the discipline that studies the principles of reasoning. It offers (1) formal languages for expressing knowledge, (2) well-understood formal semantics, and (3) automated reasoners (Antoniou & Harmelen 2004). Logic can be used by agents for making decisions and selecting courses of action. For example, if a company would like to deliver different learning content to different staff whose performances are different (e.g., sales achievement, work efficiency, peer evaluation, etc), a logic can help the learning content communicate with personal agents so that the learners are taught differently.

 

3.3     Learning Management Strategies

 

After reconstructing the e-learning content with metadata, ontologies and logic, a semantic e-learning system is technically constructed. However, a new learning management approach also needs to be constructed so that it can comply with the semantic learning environment’s requirements. From Figure 2, the learning scenarios can be divided into three stages (Huang et al. 2006). All the stages require new learning management strategies. 

 

Figure 2: Architecture of a Semantic E-Learning Framework (Huang et al. 2006)

 

The first stage is the prelearning process, which involves preparation work from both the learners and the instructors. Rather than preparing one kind of instruction for one unit of content, the instructors now are to prepare different solutions—different online multimedia learning resources, provide contextual descriptions of different learning objects for different learners, design learning paths and activities for different types of learners, and design assessments for individual sessions and whole courses. All the information will be parsed and stored into the knowledge base for future use. On the learners’ side, the intelligent agent assists learner profiling, which involves identifying learner personality and learning style by doing a series of questionnaire tests, defining learning goals and learning preferences, and clarifying personal learning responsibilities in context.

 

The second stage is the learning process, which involves various kinds of learning activities such as locating learning materials, reading materials, writing reflections, discussing with peers, self-evaluation and revision, and so on. Throughout the learning process, the intelligent agent of the learner collects real time learning data to monitor the learning progress. It uses learning signals to communicate with peer agents of other learners or with the system knowledge base against learning theories and paths in order to get adequate learning advice. Based on the learning theories and personality study results, learners with different personalities, learning styles and backgrounds are to be treated differently in different contexts; guidance will be given on an individual basis.

 

The final stage is the postlearning process, which involves reporting and evaluation of learning outcomes on both the learner and instructor sides. After each learning session or at certain checkpoints, agents could generate a learning progress report against the predefined goals and outcomes. Learning efficiency as well as the effort (e.g., time) spent on the learning activities are to be shown in the report. Further guidance for future learning path and adjustments on certain learning activities could be given if required. From the instructor’s perspective, a progress report of all involved learners from the system will provide a holistic view of the learning and teaching effectiveness in contexts, which provide concrete evidence and decision-making basis for further improvement or adjustment in learning and teaching.

 

4       Case Study: What if Developing a New Solution?

 

The assumption of the above discussion is that there has been a current non-Semantic e-learning solution, and we try to modify it. Then what if the company needs to create a new e-learning solution? In this section, Ernst and Young’s global e-learning solution case will be reviewed, and an analysis of using the idea of Semantic Web Technologies to rebuild the solution will be provided.

 

4.1     Ernst and Young’s Global E-Learning Solution Development

 

Ernst & Young Global Limited is a global leader in assurance, tax, transaction and advisory services, with about 130,000 staff helping clients retain confidence of investors, manage risk, strengthen controls and achieve potential in more than 130 countries in the world (Ernst & Young n.d.; Ernst & Young n.d.). To standardise or customise their services and make the staff members rely on explicit and tacit knowledge to solve problems, they have used the codification strategy since they frequently reuse their knowledge to achieve long-term advantage and economies of scale (Smith 2004). To support this, they needed a flexible learning system to provide a global curriculum that all the staff from different offices in the world can participate in (Werner 2002). The global curriculum focus on audit methodology, which is organised in three layers: (1) overview of the methodology, (2) detailed guidance for applying the procedures and (3) examples and leading practices. The staff’s learning about this audit methodology would be critical to its successful deployment and application. So Ernst & Young used six months to design and develop the first 300 hours of the core curriculum of the global e-learning program to initiate and support such learning, which included six main stages (Werner 2002):

 

(1)     Global Learning Committee Construction. The Committee Members were the learning leaders of Ernst & Young’s main geographic areas, who were responsible for defining the learning strategy and the development process, and approving all finished learning modules.

 

(2)       Content Creation. Ernst & Young used a modular approach to create the initial content which was divided into web-based and instructor-led learning modules based on the global audit methodology activities. Ernst & Young also assigned countries to develop the content for the modules related to a particular activity which would be bundled into logical groups later. All the modules were rated beginner, intermediate, advanced and expert.

 

(3)       Streamlined Development and Pilot Process. Once the relevant documents of the learning content had been ready, they used a streamlined development process to accelerate the actual learning module development and the build phase of the program. The Committee allocated the modules to different development teams and provided them with guidance including the detailed development process map, initial design documents, expanded design documents, leader guides, business English guide and roles description for team members (e.g. subject matter specialists, local project managers, etc.). A showcase was leveraged to test the content and gather feedback, in which approximately 80 hours of learning were delivered to the learning leaders and senior managers. The program manager and methodology team analyzed issues identified during showcase testing, and critical issues and suggestions for resolution were sent to the development teams.

 

(4)       Central Communication Point Creation. A central communication point was created to allow every developing team member from multiple countries to access the learning modules under development and make comments. Thus everyone could see what everyone else was developing.

 

(5)       Peer Review. When a learning module was developed in one country, it was systematically reviewed by a subject matter specialist from another country. Countries were asked to submit learning material related to all methodology activities, regardless of the activities their countries were assigned to develop.

 

(6)       Classroom-based Case Study Exercises Development. A separated case study team was created with a member from every country, which was responsible for creating all the information for a fictitious business. The case study was used in many of the classroom modules to create exercises to reinforce learning.

The six stages are more or less overlapped as there were different teams fulfilling responsibilities within each stage. However, the overlapped parts basically only appeared in the latter half of the whole project as the design process is actually a top-down approach (see Figure 3). The lower the levels of the teams were, the more overlapped the phases of their work were.

 

Figure 3: The Organisation Structure of Ernst & Young’s AABS Global Learning Development Project (Werner 2002)

 

4.2     Proposed Optimisation with Semantic Web Technologies

 

From Section 4.1, we can observe that Ernst and Young did not use the Semantic Web approach to construction their e-learning. Here we propose an optimisation and see how they could have construction the enterprise e-learning solution with Semantic Web idea without change the whole structure of the developing procedure. The main idea of annotating each stage is based on the three important elements of Semantic Web: metadata, ontologies and logic.

 

(1)       Global Learning Committee Construction. As the Committee Members were from different geographic areas, their contexts must be different even though their goals are similar. Hence, it is unnecessary for the Committee to use democratic voting system to decide arguable issues. Instead, once there is an argument, they can make their own decisions for their own areas with tagging different paths of subsequent procedures with their geographic labels. End-users from different areas will be provided correlative learning solutions according to their personal agents.

 

(2)       Content Creation. At this stage, Web developers and subject matter specialists should work closely together to create tagged content. That is to say, as the subject matter specialist are creating the learning content, Web developers should try to make sense of it and use tag it appropriately. Inappropriate tagging will lead to serious problems in the later stages because the metadata will be used for logic building for personal agents.

 

(3)       Streamlined Development and Pilot Process. From content creation to actual development, a critical step is to construct the ontologies, which define the all the concepts and the whole structure of all the e-learning content. And then personal agents should also be developed at the same time. Instead of using leaders and senior managers as audience to test the learning modules, end-users should be invited to test the matching between the personal agents and the e-learning content so that the internal logic can be modified according the test result.

 

(4)       Central Communication Point Creation. Besides seeing what everyone else was developing, in the central communication point, everyone can also group and integrate the productions at the same time. As different types of actions of design and development are defined, they can save much duplicate work (e.g., a figure for demonstrating the same process/theory/model may be only designed once and reused as many time in different modules as needed).

 

(5)       Peer Review. The process of peers from different countries reviewing each other’s work is actually a good opportunity to retest the logic within the communication between agents. Here there are two dimensions. One is to ensure similar input to personal agents should receive similar results. On the other hand, significantly different input to the personal agents should generate significantly different output.

 

(6)       Classroom-based Case Study Exercises Development. Each case has various characteristics that are linked to specific knowledge units. As the cases are being developed, the characteristics and the relevance to specific point of the learning content should be identified and tagged. From Table 1, as the cases’ characteristics are tagged, it is easy for personal agents to retrieve appropriate cases for specific users to learn according to how their personal knowledge structure and job description match specific tags. 

Case Number

Tag 1

Tag 2

Tag 3

Tag 4

1

X

X

 

X

2

X

 

X

 

3

 

X

 

X

4

 

X

X

X

 

Table 1: Example of Case Tagging

 

5       Conclusions

 

The main reason for incorporating Semantic Web Technologies to enterprise e-learning solution is because it will improve employees’ performance more effectively and efficiently through complying instructional design theories (e.g., First Principles of Instruction). More importantly, Semantic Web Technologies can resolve the issues existing in traditional web-based learning system (e.g., difficult to be personalised). Once a firm decides to incorporate Semantic Web technologies to their current e-learning system, they need to process the change from three dimensions: end-users, e-learning content, and management strategies. If a firm decides to create an enterprise solution to support their business processes, they can try to follow the optimised procedures of the case of Ernst and Young. In that case, on the one hand they can follow a recognised best practice to develop their own solutions. On the other hand, the new idea of Semantic Web technologies can be introduced.

 

6       References

Antoniou, G & Harmelen, F V 2004, a Semantic Web primer, MIT Press, Cambridge.

Carr-Chellman, A & Savoy, M 2004, ‘User-design research’, in D H Jonassen, Handbook of Research on Educational Communication and Technology, Lawrence Erlbaum, Mahwah, NJ, 701-716.

Dai, H & Mobasher, B 2004, ‘Integrating semantic knowledge with web usage mining for personalization’, in A Scime, Web Mining: Applications and Techniques, Idea Group, Hershey, 276-306.

Davies, J, Fensel, D, Harmelen, F V, et al. 2003, Towards the Semantic Web: Ontology-driven Knowledge Management, John Wiley & Sons, Hoboken.

Ernst & Young n.d., ‘Creating the right climate for your business success’, accessed 25/08/2008, http://www.ey.com/global/content.nsf/International/Services.

Ernst & Young n.d., ‘How we make a difference’, accessed 25/08/2008, http://www.ey.com/global/content.nsf/International/About_EY.

Gruber, T 1993, ‘A translation approach to portable ontology specifications.’ Knowledge Acquisition, vol.5, 199–220.

Hall, B 2002, ‘Six steps to developing a successful e-learning initiative: excerpts from the e-learning guidebook’, in A Rossett, The ASTD E-learning Handbook, McGraw-Hill, New York, 234-250.

Huang, W, Webster, D, Wood, D, et al. 2006, ‘An intelligent semantic e-learning framework using context-aware Semantic Web technologies.’ British Journal of Educational Technology, vol.37, no.3, 351–373.

Markellou, P, Mousourouli, I, Spiros, S, et al. 2005, ‘Using Semantic Web Mining Technologies for Personalized e-Learning Experiences’. in Proceedings of Proceedings of the web-based education, Grindelwald, Switzerland.

Merrill, M D 2002, ‘First principles of instruction.’ Educational Technology, Research and Development, vol.50, no.3, 43-59.

Mobasher, B, Jin, X & Zhou, Y 2003, ‘Semantically enhanced collaborative filtering on the web.’ EWMF, 57-76.

Smith, A D 2004, ‘Knowledge management strategies: a multi-case study.’ Journal of knowledge management, vol.8, no.3, 6-16.

Strother, J B 2002, ‘An Assessment of the Effectiveness of e-learning in Corporate Training Programs.’ International Review of Research in Open and Distance Learning, vol.3, no.1.

Vygotsky, L S 1978, Mind in Society, Harvard University Press, Cambridge.

Werner, T 2002, Best practices for e-learning: top entries in the best practices category Brandon-Hall, Sunnyvale, CA


本页浏览次数:8851

发表评论

电子邮件地址不会被公开。 必填项已用*标注


*