Copyright Emerald Group Publishing Limited 2008 Web mining applications in e-commerce and e-services
Edited by I-Hsien Ting
Introduction
With the flourishing of the web, the use of web- and internet-based technology seems to be ubiquitous in education and training. During recent decades the most important innovations in educational systems have been related to the introduction of new technologies as part of web-based education. This mode of education has grown considerably in importance, and thousands of web courses have been deployed in the past few years. However, many of the current web-based courses are built around static learning materials, which do not take into account the diverse learning needs of students.
In a traditional educational setting the educators can obtain feedback directly from students through face-to-face interaction and behaviour analysis ([25] Sheard et al. , 2003). However, in the electronic learning environment it is impossible to have this feedback. Adaptive and intelligent web-based educational systems have been seen as a solution for enriching the learning environment for the individual. Most past research on web-based learning focused on the issues of adaptive presentation, adaptive navigation support, curriculum sequencing, and intelligent analysis of a student's behaviour. In order to offer learners personalised education data mining technologies are being adopted. Data mining or knowledge discovery in databases (KDD) is the automatic extraction of implicit and interesting patterns from large data collections ([10] Klosgen and Zytkow, 2002). KDD can mine some useful information from learning portfolios and use this information for learning process modelling and student modelling ([29] Tang and McCalla, 2002).
The paper begins with a review of web mining in e-learning. Then we introduce some issues of web mining in VLE, including a classification of user behaviour in VLE for web mining, interaction hierarchical model and general architecture. Finally, design and application of experimental prototypes is addressed.
Review of web mining in e-learning
Web mining can be broadly defined as the discovery and analysis of useful information from the worldwide web (WWW). [5] Cooley et al. (1997) present the taxonomy of web mining, i.e. web content mining and web usage mining. Web content mining focuses on searching for user-desired information online using traditional search engines, such as Lycos, Ala Vista, WebCrawler and ALIWEB. In recent years, some research has been focused on developing intelligent tools for information retrieval, adopting either an agent-based or database approach. Discovery of user behaviour patterns on the web is the main task of web usage mining. In order to mine useful information, two types of tools have been developed, i.e. patterns discovery tools and analysis tools ([5] Cooley et al. , 1997).
According to the taxonomy of web mining, web mining in e-learning can be classified according to content mining or usage mining. But there are some important differences between web mining in e-commence and web mining in e-learning ([20] Romero and Ventura, 2007):
- Domain . The e-commerce purpose is to guide clients in purchasing while the e-learning purpose is to guide students in learning ([21] Romero et al. , 2004).
- Data . In e-commerce the web mining data sources are normally simple web server access logs, but in e-learning there is more information coming from the user's education interaction log ([18] Pahl and Donnellan, 2003). The user-modelling paradigm is also different in both systems.
- Objective . The objective of web mining in e-commerce is to increase profit, which is tangible and can be measured in terms of amounts of money, number of customers and customer loyalty. However, the objective of web mining in e-learning is to improve the learning. This goal is more subjective and more subtle to measure.
- Techniques . Education systems have special characteristics that require a different treatment of the mining problem. As a consequence, some specific data mining techniques are needed to address the process of learning in particular ([11] Li and Zaïane, 2004). Some traditional techniques can be adapted, some cannot.
Unlike traditional teaching environments, when students work in electronic environments, it is impossible for the educator to retrieve students' feedback information. Educators must look for other ways to gain this information. Web-based education systems can normally record the learner's accesses in web logs that provide a raw trace of the learners' navigation on the site. There are several types of logs: server log file, client log file, and proxy log file ([19] Pierrakos et al. , 2003):
- Server log file . This constitutes the most widely used data source for performing data mining, containing just the bare details of timing, path and input-response. Unfortunately, there is a single log file for all students.
- Client log file . This consists of a set of log files, one per student, and contains information about the interaction of the user with the system.
- Proxy log file . This consists of a set of log files of caching between client browsers and web servers. This information complements the server log file.
Using these log files, web-based learning environments are able to record most learning behaviours of the students, and hence are able to provide a huge amount of learning profile ([20] Romero and Ventura, 2007). Because they have different data sources and objectives, it is necessary to deal separately with web mining applications in different types of e-learning systems. [20] Romero and Ventura (2007) distinguish between three different types of web-based education systems: web-based courses, well-known learning content management systems, and adaptive and intelligent web-based educational systems.
Web-based course
Web-based courses refer to the courseware that uses static or dynamic web page technology. No matter what content they present, they have the same data source ([26] Srivastava et al. , 2000). The pages of web-based courses usually consist of texts, graphics, videos, sounds, etc. The inter-page structure information is hyper-links connecting one page to another. Intra-page structure information can be represented as a structured tree, where the HTML tag is the root.
Web mining can be used to discover how students use the course, how a pedagogical strategy affects different types of students, in which order the students study sub-topics, the pages/topics that students skip, and how much time the students spend on a single page, a chapter or the full course.
Learn content management system
Learning content management systems (LCMS) are platforms that offer a great variety of channels and workspaces to facilitate information sharing and communication between participants in a course. There are many commercial or free LCMS, such as Blackboard, WebCT, Moodle, aTutor, etc. These systems accumulate large amounts of log data related to students' activities, such as reading, writing and even communicating with peers. An LCMS also provides a database that stores all the system's information: personal information regarding the user (profile), academic results, and user interaction data.
Web mining can be applied to explore, visualise and analyse data in order to identify useful patterns and to evaluate web activity to derive more objective feedback on one's teaching and to learn about how students learn through the LCMS ([22] Romero et al. , 2007; [27] Talavera and Gaudioso, 2004).
Adaptive and intelligent web-based educational systems
Adaptive and intelligent web-based educational systems (AIWBES) provide an alternative to the traditional just-put-it-on-the-web approach in the development of web-based educational courseware ([3] Brusilovsky and Peylo, 2003). The data from AIWEBS are semantically richer and can lead to more diagnostic analysis than data from a traditional web-based education system ([13] Merceron and Yacef, 2004). AIWBES attempt to be more adaptive by building a model of the goals, preferences and knowledge of each individual student and using this model throughout the interaction with the student in order to adapt to the needs of that student.
Web mining can be used in order to discover the causes of problems in the system, for example, incorrect feedback statements ([16] Nilakant and Mitrovic, 2005), to adapt the level to the progress of the learner ([21] Romero et al. , 2004), to suggest personalised learning experiences and activities for the students ([28] Tang and McCalla, 2005).
In this paper, we focus on web mining applications in the virtual learning environment (VLE), which is similar to web mining in AIWBES. Next, we will describe some detail issues of web mining in VLE.
Interest mining in VLE
The virtual learning environment (VLE) is a learning environment where the necessary interaction and collaboration takes place in the virtual world of web and Internet-based applications. The main difference between a VLE and other web-based learning environments is the possibility of communication and collaboration with peers and tutors within the same virtual environment that holds the content.
VLE is not restricted to systems including some 3D/virtual reality technology ([6] Dillenbourg, 2000), but in this paper we focus on the Web3D VLE. The use of Web3D in education presents many of the advantages with respect to traditional learning practice. First, it provides a wide range of rich experiences, some of which are impossible to experience in the real world because of cost, danger, or for some other reason. Second, the use of 3D graphics can allow a more realistic and detailed presentation compared to 2D technology ([4] Chittaro and Ranon, 2007). There are many successful cases of Web3D ([2] Brenton et al. , 2007; [8] Hamalainen et al. , 2006; [9] Ieronutti and Chittaro, 2007; [14] Morphew, 2000; [15] Mzoughi et al. , 2007; [30] Wyeld, 2005). Under the support of the National "Ten-Five" Key Technologies R&D Project of China, Zhejiang University provides a virtual chemical laboratory and circuit laboratory for undergraduate students ([7] Gu et al. , 2005; [17] Ouyang et al. , 2005; [31] Yabo and Miaoliang, 2002).
Despite the advances in VLEs there still exist many problems to be solved. The first one is system consistency. Consistency focuses on the characteristic that the VLE system maintains one consistent state of environment (including user avatar) simultaneously. Hence, the VLE system must distribute the state update message to all other participants continuously. With growth in the number of participants and the complexity of the environment, the number of state update messages will increase explosively, which brings the network massive traffic and low system performance. This results from neglecting the learner's implicit interest, which will lead to learner cognition overload, thus generating learning difficulty.
According to research in psychology, any interaction in VLE has implications for user implicit interest. Hence, we propose the methodology of mining the learner's implicit interest via explicit behaviour analysis. Like other information retrieval applications (IRA) ([1] Baeza-Yates and Ribeiro-Neto, 1999), VLE needs to mine students' requirements, searching for or filtering useful information within a huge amount of data, and then provide students with the information they need. Therefore, web mining in e-learning systems, especially in VLE, is essential.
The objective of interest mining in VLE is to find the student's implicit requirement, such as user interest. The data sources for interest mining come from user interaction in VLE. According to general web mining methodology in other application areas, the web mining application in VLE consists of two modules: a behaviour retrieval sub-system (termed BR) and an interest extraction sub-system (termed IE). The BR module consists of low-level signal capture, data clustering and classification; the IE module is charged with association rule mining and sequential pattern mining. Design of these two modules is based on the interest model in VLE.
Interaction model in VLE
Humans are highly social creatures, so it is crucial for us to be able to perceive what others are doing and to infer the implicit meaning from their gestures and expressions. Hence, user interaction in VLE, especially social interactions, are important cues for mining user interest. This section outlines the concept and the classification mode of interactions in VLE.
The term "interaction" follows the definition provided by [12] Manninen (2001): "... interaction set consisting of a large number of individual action and interaction types and possibilities that allows more complex interaction sequences". Complexity refers to the more natural forms of rich interaction. It seems that the computer's virtual counterpart lags behind the real-world one, but we can cluster and classify these interactions according to the model shown in Figure 1 [Figure omitted. See Article Image.]. The model consists of five main catalogues: travel, gaze, manipulate, gesture, and conversion:
Travel . The interaction is between user and environment, which is based on the user's location update. User travel in scene refers to user's intention to explore undiscovered areas in virtual world according to direction.
Gaze . The interaction is between two users or between user and environment. User gaze refers to user's intention to examine the detail of object according to user's view.
Manipulate . The interaction is between user and environment objects. According to inner interest in accomplishing tasks, the user will manipulate the object, push box, pick up a tool, turn on television, etc.
Gesture . The interaction is social interaction between two users. A user makes different expressive movements of parts of the body, especially the hands or head, according to the specific intention. For example, a student bows to his tutor to show respect.
Talk . This is language-based social interaction between two users via speech or text/image message.
The interaction hierarchical model in Figure 2 [Figure omitted. See Article Image.] shows the process for mining implicit interest via explicit interaction in VLE. According to the different interaction targets, interaction in VLE can be divided into two sub-models: user-object interaction sub-model shown in (1), and user-user social interaction sub-model shown in (2). Each sub-model indicates two interaction layers, i.e. explicit interaction and implicit interaction. The explicit interaction layer covers explicit interaction information retrieval, and implicit interaction layer is used for mining implicit interest, i.e. intention and interest. Although there are five types of interaction in VLE, the retrieval of explicit interaction information has three steps: collecting signal at sensor sub-layer, clustering and classification, detecting outline at pattern sub-layer and generating output at goal sub-layer.
The main idea of this structure is to divide and classify interactions in VLE. Creating a hierarchical structure starts from low-level signal-type actions and ascends to the level at which the cognitively-generated goals and objectives define the purpose of the interaction itself.
Based on the interaction clustering and classification in VLE, we propose that user-implicit interaction information can be extracted from low-level explicit interaction information. Next, we introduce details of interest mining in VLE.
General architecture for interest mining in VLE
Here we introduce a general architecture for interest mining in VLE. According to the interaction hierarchical model, general architecture divides the interest mining process into two main parts: behaviour retrieval (called BR) and interest extraction (called IE). Behaviour retrieval includes data pre-processing, pattern discovery and interaction goal mining. The interest extraction part includes knowledge-based data mining technology as the interest-mining engine. The general architecture for interest mining in VLE is shown in Figure 3 [Figure omitted. See Article Image.].
Data collection
The first step in the VLE interest mining process is the collection of original signal data, which can be captured and analysed to provide useful information about the user behaviour. Although there are three data sources for web mining (server-side data, client-side data and proxy-side data) ([19] Pierrakos et al. , 2003), the main data source for interest mining in VLE is client-side interaction data, which can record user behaviour or explicit user input.
Client side data are captured from the local host, which accesses the VLE site. Traditional technologies for acquiring client data involve dispatching a remote agent, implemented in Java or JavaScript ([23] Shahabi et al. , 2001; [24] Shahabi et al. , 1997). However, these technologies require the authorisation of the user, who may activate the local security mechanisms for restricting the operation of Java and JavaScript programs from their browsers. Thus, the user's behaviour or explicit input data can be captured by local signal sensors, which transfer these original data to pre-processing applications.
Data pre-processing
Data collection is the first phase of interest mining in VLE. Because the collected original signal data tend to be incomplete, noisy and inconsistent, these raw data must be assembled into a consistent, integrated and comprehensive view for next pattern discovery. As the most essential phase of interest mining in VLE, data pre-processing routines attempt first to fill in missing values, smooth out noise while identifying outliers and correct inconsistencies in data, then integrate and transform data into forms appropriate for pattern discovery.
The data pre-processing step is to some extent domain-dependent; that is, the value and the sequence of user behaviour affect the decision about which data are relevant. It is also strongly dependent on the type and the quality of the data
Pattern discovery
In the pattern discovery stage machine learning and other statistical methods are used to extract patterns of user interaction from the pre-processed clean signal data. Similar to most data mining applications, web mining adopts four approaches for pattern discovery (clustering, classification, association discovery and sequential pattern discovery). The majority of methods for interest mining in VLE are clustering methods.
In this study pattern discovery aims to discover implicit interaction data from a large amount of pre-processed signal data. Among data mining clustering techniques the model-based clustering method is particularly well-known; it can discover characteristic descriptions for each pre-processed signal data group, where each group represents an interaction class.
Interaction goal and implicit interest mining
In contrast to pattern discovery, interaction goal and implicit interest mining adopt classification methods. The goal of classification is to identify the distinguishing characteristics of predefined goals or interests, based on a set of instances. These characteristics can be used both for understanding the existing data and predicting how the new instance will behave.
In the context of interest mining in VLE, classification methods can be used for modelling user behaviour goal characteristics and implicit interest.
Design and application of experimental system
We have designed and developed a prototype system and experimented with our interest mining approach on users' gesture signal data collected from a desktop VR system, using 5DT CyberGlove system and a 6-degree of freedom (6DOF) position tracker. We evaluated our data pre-processing, pattern discovery, interaction goal and implicit interest mining methods.
Interest mining flowchart
The flowchart of interest mining in an experimental prototype is shown in Figure 4 [Figure omitted. See Article Image.].
In the experimental prototype, the user behaviour signal (e.g. gesture) is captured by CyberGlove system; a 6DOF position tracker captures the user's hand position and motion signal simultaneously. These two original signal data are cleaned, integrated and transformed in data pre-processing modules. By searching the gesture feature in a characteristic database, the system clusters the pre-processed data and finishes pattern discovery. Then goal mining and interest mining modules accomplish the user implicit interest extraction via gesture. Finally, the system updates the user interest state, which can be used to promote system performance.
Application of user interest
The main application of user interest in experimental prototypes is filtering network traffic between users. As we mentioned before, each client in VLE needs to distribute a state update message to all other participants continuously, which leads to explosive network traffic. After mining the user's implicit interest, the system can transfer the local state update message to different participants in a different frequency. Moreover, the system can control the level of detail (LOD) of the state update message to cut down network traffic, i.e. the remote participant, in whom the user is more interested, can receive more detailed information from the user than other clients receive.
Prototype system architecture
System architecture is the main factor of prototype system design. Once the local interest state is updated, the client must inform other clients as soon as possible. The update message propagates via the channel determined by system topology. Generally, there are three basic system topologies and three mutant topologies, which are shown in Figure 5 [Figure omitted. See Article Image.].
Centralised topology (a) has the advantage of easily providing a filtering mechanism at the server side (core node). However, the scalability is limited because of the server-side bottleneck. Distributed topology (b) provides the opposite approach. There is no central server as well as server-side bottleneck, for each client (termed "peer") is responsible for message transmission. Nevertheless, it is difficult to implement the data filter mechanism. The hybrid topology is the compromise plan of building on the advantages of the two previously mentioned architectures. In Figure 5 [Figure omitted. See Article Image.], (c)-(f) belong to this topology, which solves the difficulty of (a) and (b), but they still have the problem of high cost due to multiple servers and additional information exchange among servers ([32] Yu and Choy, 2001).
Our experimental prototype adopts hybrid architecture shown in Figure 6 [Figure omitted. See Article Image.]. Each client (user peer) consists of several major modules, such as communication module, message module, interest management module and application module. The server (super peer) has additional modules for system management, for example administrator sub-module:
- Communication module . The communication module is in charge of socket channel management. According to the difference of payload type, the communication module can use communication port range from reliable socket for unicast to unreliable socket for multicast.
- Message module . This part generates and parses the message for transfer between peer and peer or between peer and server. Another function of the message module is to control the level of detail (LOD) of information; that is, the module will generate a different copy of the same user action according to different interests.
- Interest mining module . The interest mining module has the role of monitoring user behaviour (especially interaction), mining user interest; updating user local correlation graph and maintaining global correlation at the server side.
Summary
In this paper, an approach to user interest mining has been presented for the automatic retrieval and analysis of user interaction in the VLE. Web mining in education has already produced a number of interesting results, including the intelligent education system and the general content search engine. However, research about user interest mining based on interaction analysis is rare. This paper described work in progress, and the next stage includes development of a demonstration system for testing.
Proceedings of the 9th IEEE International Conference on Tools with Artificial Intelligence, Newport Beach, CA, IEEE Computer Society, New York, NY, 5-8 November
EUN Conference 2000, Learning in the New Millennium: Building New Education Strategies for Schools, Workshop on Virtual Learning Environments
ITHET 2005: 6th International Conference on Information Technology Based Higher Education and Training
International Conference on Computational Science 2005
Proceedings of the Congress E-learning, Montreal, Canada
Proceedings of the 7th IEEE Intlernational Workshop on Research Issues in Data Engineering (RIDE)
Workshop on Artificial Intelligence in CSCL, 16th European Conference on Artificial Intelligence
Proceedings of the 18th National Conference on Artificial Intellegience (AAAI-2002), Edmonton, Canada
Proceedings of the 5th IEEE International Conference on Advanced Learning Technologies, ICALT 2005, Kaohsiung, Taiwan
| [Reference] |
| 1. Baeza-Yates, R. and Ribeiro-Neto, B. (1999), Modern Information Retrieval, Wesley-Longman, Harlow. |
|
| 2. Brenton, H., Hernandez, J., Bello, F., Strutton, P., Purkayastha, S., Firth, T. and Darzi, A. (2007), "Using multimedia and Web3D to enhance anatomy teaching", Computers and Education, Vol. 49 No. 1, pp. 32-53. |
|
| 3. Brusilovsky, P. and Peylo, C. (2003), "Adaptive and intelligent web-based educational systems", International Journal of Artificial Intelligence in Education, Vol. 13, pp. 156-69. |
|
| 4. Chittaro, L. and Ranon, R. (2007), "Web3D technologies in learning, education and training: motivations, issues, opportunities", Computers and Education, Vol. 49 No. 1, pp. 3-18. |
|
| 5. Cooley, R., Mobasher, B. and Srivastava, J. (1997), "Web mining: information and pattern discovery on the world wide web", , pp. 558-67. |
|
| 6. Dillenbourg, P. (2000), "Virtual learning environments", , available at: tecfa.unige.ch/tecfa/publicat/dil-papers-2/Dil.7.5.18.pdf. |
|
| 7. Gu, R., Zhu, M., Dong, Y., Shi, D. and Wang, Y. (2005), "A case study of virtual circuit laboratory for undergraduate student courses", , Vol. 2005, pp. 21-4. |
|
| 8. Hamalainen, R., Manninen, T., Jarvela, S. and Hakkinen, P. (2006), "Learning to collaborate: designing collaboration in a 3-D game environment", The Internet and Higher Education, Vol. 9 No. 1, pp. 47-61. |
|
| 9. Ieronutti, L. and Chittaro, L. (2007), "Employing virtual humans for education and training in X3D/VRML worlds", Computers and Education, Vol. 49 No. 1, pp. 93-109. |
|
| 10. Klosgen, W. and Zytkow, J. (2002), Handbook of Data Mining and Knowledge Discovery, Oxford University Press, New York, NY. |
|
| 11. Li, J. and Zaïane, O. (2004), "Combining usage, content, and structure data to improve web site recommendation", International conference on e-commerce and web technologies, pp. 305-15. |
|
| 12. Manninen, T. (2001), "Rich interaction in the context of networked virtual environments - experiences gained from the multi-player games domain", in Blanford, A., Vanderdonckt, J. and Gray, P. (Eds), Joint Proceedings of HCI 2001 and IHM 2001 Conference, Springer-Verlag, New York, NY, pp. 383-98. |
|
| 13. Merceron, A. and Yacef, K. (2004), "Mining student data captured from a web-based tutoring tool: initial exploration and results", Journal of Interactive Learning Research, Vol. 15 No. 4, pp. 319-46. |
|
| 14. Morphew, V.N. (2000), "Web-based learning and instruction: a constructivist approach", in Lau, L. (Ed.), Distance Learning Technologies: Issues, Trends and Opportunities, Idea Group, London, pp. 1-15. |
|
| 15. Mzoughi, T., Herring, S.D., Foley, J.T., Morris, M.J. and Gilbert, P.J. (2007), "WebTOP: a 3D interactive system for teaching and learning optics", Computers and Education, Vol. 49 No. 1, pp. 110-29. |
|
| 16. Nilakant, K. and Mitrovic, A. (2005), "Application of data mining in constraint-based intelligent tutoring systems", in Looi, C.-K., McCalla, G., Bredeweg, B. and Breuker, J. (Eds), Proceedings of the 12th International Conference on Artificial Intelligence in Education AIED, Amsterdam, 18-22 July, IOS Press, Washington, DC, pp. 896-8. |
|
| 17. Ouyang, Y., Dong, Y., Zhu, M., Huang, Y., Mao, S. and Mao, Y. (2005), "ECVlab: a web-based virtual laboratory system for electronic circuit simulation", , pp. 1027-34. |
|
| 18. Pahl, C. and Donnellan, C. (2003), "Data mining technology for the evaluation of web-based teaching and learning systems", . |
|
| 19. Pierrakos, D., Paliouras, G., Papatheodorou, C. and Spyropoulos, C.D. (2003), "Web usage mining as a tool for personalization: a survey", User Modelling and User-Adapted Interaction, Vol. 13 No. 4, pp. 311-72. |
|
| 20. Romero, C. and Ventura, S. (2007), "Educational data mining: a survey from 1995 to 2005", Expert Systems with Applications, Vol. 33 No. 1, pp. 135-46. |
|
| 21. Romero, C., Ventura, S. and Bra, P.D. (2004), "Knowledge discovery with genetic programming for providing feedback to courseware authors", User Modeling and User-Adapted Interaction, Vol. 14 No. 5, pp. 425-64. |
|
| 22. Romero, C., Ventura, S. and Garcia, E. (2007), "Data mining in course management systems: Moodle case study and tutorial", Computers and Education, Vol. 49, pp. 1-17. |
|
| 23. Shahabi, C., Banaei-Kashani, F. and Faruque, J. (2001), "A reliable, efficient, and scalable system for web usage data acquisition", paper presented at WebKDD'01 WorkShop in conjunction with ACMSIGKDD 2001. |
|
| 24. Shahabi, C., Zarkesh, A.M., Abidi, J. and Shah, V. (1997), "Knowledge discovery from user's web-page navigation", , pp. 20-9. |
|
| 25. Sheard, J., Ceddia, J., Hurst, J. and Tuovinen, J. (2003), "Inferring student learning behaviour from website interactions: a usage analysis", Education and Information Technologies, Vol. 8 No. 3, pp. 245-66. |
|
| 26. Srivastava, J., Cooley, R., Deshpandé, M. and Tan, P.-N. (2000), "Web usage mining: discovery and applications of usage patterns from web data", ACM SIGKDD Explorations Newsletter, Vol. 1 No. 2, pp. 12-23. |
|
| 27. Talavera, L. and Gaudioso, E. (2004), "Mining student data to characterize similar behavior groups in unstructured collaboration spaces", , pp. 17-23. |
|
| 28. Tang, T. and McCalla, G. (2005), "Smart recommendation for an evolving e-learning system", International Journal on E-Learning, Vol. 4 No. 1, pp. 105-29. |
|
| 29. Tang, T.Y. and Mccalla, G. (2002), "Student modeling for a web-based learning environment: a data mining approach", , pp. 967-8. |
|
| 30. Wyeld, T.G. (2005), "The pedagogical benefits of remote design collaboration in a 3D virtual environment: a case study", , IEEE Computer Society, New York, NY, 5-8 July 2005, pp. 824-6. |
|
| 31. Yabo, D. and Miaoliang, Z. (2002), "Web-based VR-form virtual laboratory", Chemical Engineering Education, Vol. 36 No. 2, pp. 102-7. |
|
| 32. Yu, S.-J. and Choy, Y.-C. (2001), "A dynamic message filtering technique for 3D cyberspaces", Computer Communications, Vol. 24 No. 18, pp. 1745-58. |
|
| [Appendix] |
| Corresponding author |
| Rong Gu may be contacted at: gr@zjut.edu.cn |
| |
| [Author Affiliation] |
| Rong Gu, College of Computer Science and Technology, Zhejiang University, Hangzhou, China |
| |
| Miaoliang Zhu, College of Computer Science and Technology, Zhejiang University, Hangzhou, China |
| |
| Liying Zhao, College of Educational Science and Technology, Zhejiang University of Technology, Hangzhou, China |
| |
| Ningning Zhang, College of Educational Science and Technology, Zhejiang University of Technology, Hangzhou, China |
| [Illustration] |
| Figure 1: Catalogue of interaction in VLE |
| |
| Figure 2: Interactions hierarchical model |
| |
| Figure 3: General architecture for interest mining in VLE |
| |
| Figure 4: Flowchart of interest mining in experimental prototype |
| |
| Figure 5: System architecture (S: server; C: client) |
| |
| Figure 6: Experimental prototype architecture |
| |
My Research