Research on Learning Resources Organization Model 55 3.2.4 Learning Content Scheduling Algorithm The scheduling algorithm of learning contents explains how to choose a appropriate learning object when the learner begins to learn a knowledge cluster. At first, it de- termines the knowledge point according to the sequence relations of knowledge, and then chooses the best learning object according to the learner’s learning style and the learning context. The organization of learning contents mainly includes two aspects. One is how to show the appropriate learning contents after choosing the knowledge cluster to learn, the other is how to diagnose difficult knowledge points and choose the appropriate learning content when the learner ends the learning and does not pass the test. Fig 5 shows the relation of knowledge points sequence relations and person- alized manifestation. Fig. 5. The relation of knowledge points sequence relations When the learner begins to learn a new knowledge cluster, the system runs the scheduling algorithm as follows. (1) The learner chooses a knowledge cluster according to his needs. (2) The system provide the learner with pre-test according to his learning state and the selected knowledge cluster, in order to test whether the learner has the ability to learn the knowledge cluster. (3) Providing the learner with the best appropriate knowledge point on the basic of knowledge points’ sequence relations. Choosing knowledge points is the strategy of presenting the learning object according to the learner’s learning style, personalized characteristic, etc. (4) Providing the learner with the appropriate learning object from the same knowledge. (5) Repeating (3), (4) , and providing all the knowledge points in the knowledge cluster to the learner step by step. When the learner finishes the learning of a knowledge cluster, but he does not pass the test, the system will find the difficult knowledge point, and then choose the ap- propriate learning object of the knowledge point according to the learner’s personal- ized characteristic. 56 Q. Liu and Z. Sun 3.3 Post-test and Evaluation The result of examination may be changed with different environment or psychologi- cal condition, so it is inaccurate to judge his master degree depending on the result of examination. Therefore, we should consider the examination result, and also we should consider the learner’s learning activity, such as participating in discussion, answering the questions and so on during the evaluation, teachers can evaluate the learner according to the information of his learning activity. The system can get the final result by considering the examination and the teachers’ evaluation and provide the grade to the learner. 4 Application of Learning Resources Organization Model We implement our algorithm in our system named E-learning Service Platform (ELSP), which has been developed with J2EE. ELSP is a complex e-learning system included learning management system, content management system, testing system and so on. According to the content organization algorithm, the related databases include information of the learner, attributes of knowledge, attributes of knowledge cluster and learning state and style and so on. In which, the learner’s information is comprised of name, sex, birthday, password, ID, telephone and email etc. Knowledge point attributes include name, up-node, sub-node, cognitive rank, etc. The learning state and style includes name, learned knowledge cluster, ability level, style, attempt times, test time, difficult knowledge point and the progress of learning and so on. At same time, we design e-learning courseware named “database theory and technology” with a lot of learning objects. The courseware is used for college students in Educa- tion Technology in Central China Normal University. The knowledge tree is pre- sented in Fig 6 based on database. Fig. 6. The dynamic knowledge tree The following information in fig7 shows the learner’s progress, the level of learned knowledge cluster, attempt times and learning time. It reflects the learning situation dynamically. Research on Learning Resources Organization Model 57 Name Learned Knowledge Cluster Mastering Level Trying times Testing time Difficult knowledge point Next Knowledge point Zhang min Requirement analysis Better 1 2006-3-20 Null Design of concept. Wang fang Design of concept. Bad 1 2006-4-20 View integration Logical structure Li hai Design of concept. Good 3 2006-4-19 Null Logic structure Liu ruibing Logical structure Good 1 2006-4-26 View integration Physical design. …… …… …… … … …… …… …… Fig. 7. Organization and schedule of learning resources Compared to Simple Sequence Specification, Learning Design Specification, our designed algorithm focuses on the organization and schedule of learning resources according to different learners and learning situation. It mainly has the following merits: (1) It can dynamic construct the learning content, and provide different content to learners, which met the personalized needs better. (2) The resources described with e-learning technology standards can realize the resource to be reused and shared. (3) During the learning process, the algorithm is checking learning situation and learners’ state to schedule learning resources, so its performance is better, its effi- ciency is higher. 5 Conclusions The paper constructs the learning resources framework model, defines granularity of the learning object, designs the learning content scheduling algorithm, which includes pre-test, knowledge points learning, post-test and evaluation. The paper put forward the learning content scheduling algorithm from knowledge sequencing and personal- ized representation as two dimensions. the following work mainly includes: (1) the learning activities as an important origin of the testing data, needs the further design- ing and developing. (2) how to obtain the information of learner’s personalized char- acteristic needs to have further research. (3) the algorithm of personalized presenting the learning object needs to perfect. Acknowledgments. The paper is supported by Specialized Research Fund for the Doctoral Program of Higher Education, Ministry of Education of China (NO20050511002), supported by the Programme of Introducing Talents of Discipline to Universities Ministry of Education and State Administration of Foreign Experts Affairs of China (NOB07042), Supported by NSFC of China(NO60673094) , partly supported by the Natural Science Foundation of Hubei Province(NO.2006ABC011) 58 Q. Liu and Z. Sun and by National Great Project of Scientific and Technical Supporting Programs Funded by Ministry of Science & Technology of China During the 11th Five-year Plan (NO. 2006BAH02A24). References 1. Wiley, D.A., et al.: The Instructional Use of Learning Objects, http://www.reusability.org (2001) 2. Abruf: IEEE Learning Technology Standards Committee (LTSC) Learning Object Meta- data - Draft Document v6. 0 (2001), http://ltsc.ieee.org/doc/wg12/ 3. IMS Simple Sequencing Best Practice and Implementation Guide (2006), http://www. imsglobal.org/simplesequencing/ssv1p0/imsss_bestv1p0.html 4. IMS Global Learning Consortium: Learning Design Specification (2007), http://www. imsglobal.org/learningdesign/ 5. The SCORM Overview, SCORM version 1.2., http://www.adlnet.org/ ADLDOCS/Doeuments/SCORM-1.2-Overview.pdf 6. Issack Santally, M.: A Learning Object Approach to Personalized Web-based Instruction. European Journal of Open Distance and E-learning (EURODL) 7. Furugori, N., et al.: COALE: Collaborative and Adaptive Learning Environment (2007), http://citeseer.ist.psu.edu/509287.html 8. lu, W., hui, Y., le, D.: Research on knowledge representation and illation based on fuzzy relation data model of ITS. Journal of E-education Research 4, 30–34 (2005) 9. Xiaoyong, H., Zhiting, Z.: The taxonomy of reused learning objects. Journal of E- education Research 8, 10–15 (2003) 10. Steinacker, A., Seeberg, C., Reichenberger, K., Fischer, K., Steinmetz, R.: Dynamically Generated Tables of Contents as Guided Tours in Adaptive Hypermedia Systems. In: Pro- ceedings of the EdMedia & EdTelecom (1999) 11. Fischer, S., Steinmetz, R.: Automatic Creation of Exercises in Adaptive Hypermedia Learning Systems. In: Proceedings of 10th ACM conference on Hypertext (HT 2000) (to appear, 2000) 12. Cisco: Reusable Learning Object Strategy, http://www.cisco.corn/warp/ public/10/wwtraining/elearning/imple-ment/rlo-stratrgy.pdf F. Li et al. (Eds.): ICWL 2008, LNCS 5145, pp. 59–68, 2008. © Springer-Verlag Berlin Heidelberg 2008 Web Contents Extracting for Web-Based Learning Jiangtao Qiu 1, 2 , Changjie Tang 2 , Kaikuo Xu 2 , and Qian Luo 2 1 School of Economic Information Engineering, South Western University of Finance and Economics, Chengdu, China 610074 2 Computer School, Sichuan University, Chengdu, China 610065 jiangtaoqiu@gmail.com Abstract. Web mining has been applied to improve web-based learning. Con- tent-based Web mining usually focuses on main contents of web page. This paper proposes a novel approach to automatically extract main contents from web pages. Compared with existed studies, the method may determine whether a web page contains main contents, and then extracts main contents without using DOM-Tree and template. Main contributions include: (1) Introducing a new concept of Block and proposing a method to partition web page to blocks. Main contents and noise contents may be well partitioned into different blocks. (2) Introducing a concept of Web Page Block Distribution and studying its feature. Based on Block Distribution, we may effectively determine whether the web page contain main contents, and then extract main contents via outlier analysis. Experiments demonstrate utility and feasibility of the method. Keywords: Web-Based Learning, Web Contents Extracting, Web Mining. 1 Introduction Web mining has been applied to improve web-based learning. Content-based Web mining usually focuses on main contents of web page, and regards advertisements, navigation sidebar and copyright notice, etc as noise. Noise has a negative impact on analysis of web contents, for example, reducing accuracy of web classification and clustering. Some studies [1, 2] analyzed how noises harm web mining. In Content-based web mining, usually, it needs to collect web pages, and then de- termine whether web pages include main contents. We refer having main content page and not having main content page to content page and non-content page respectively. We need to extract main contents from content pages. The existed methods have some limits on applications: (a) all web pages are assumed to be content pages. This as- sumption, however, is indefensible in application. In most cases, the collected web pages are mixture of content pages and non-content pages. (b) template-based methods assume that web pages coming from same web site have same layout. Hence they may correctly extract contents from web pages matching with template. To get correct contents from all web pages, however, it is necessary to build templates for all involved website. This is a very arduous work. (c) DOM-Tree-based method firstly converts a piece of web pages to a Dom-Tree, then extract contents from the DOM-Tree. Refer to our naive purpose that get main contents from web pages, Dom-Tree, with complicated structure and rich functions, is obviously cost-expensive. . Web-Based Learning, Web Contents Extracting, Web Mining. 1 Introduction Web mining has been applied to improve web-based learning. Content-based Web mining usually focuses on main contents of. method may determine whether a web page contains main contents, and then extracts main contents without using DOM-Tree and template. Main contributions include: (1) Introducing a new concept. knowledge point on the basic of knowledge points’ sequence relations. Choosing knowledge points is the strategy of presenting the learning object according to the learner’s learning style, personalized