【作者】 陈清华；

【导师】 申瑞民；

【作者基本信息】 上海交通大学 ， 计算机软件与理论， 2008， 硕士

【摘要】 E-Learning以网络技术和多媒体技术为核心,给教师和学生、学生和学生之间搭建了一座通信桥梁,为更多渴望知识的人提供获取知识的一种新的手段。然而,网络上知识载体呈现爆炸增长趋势,人们面临着“信息爆炸,但知识贫乏”的尴尬。同时,与传统教育方式相比,网络教育平台上的学员们常常因缺乏有效的指导而导致学习效率低下,因缺乏同伴间的相互交流、相互学习,存在着前所未有的“孤独感”。目前,对于这些问题的相关研究屡有创新的进展。从基于内容的过滤到协同过滤,到自组织社区;中国网络教育技术标准化的从无到有;Ontology的引入及其应用等,每深入一步研究,都有新的发现。研究者们意识到:个性化服务若单方面地考虑用户本身的历史数据作为分析源头,其准确率在某些环境下并不适用;通过横向有效地挖掘和利用相互间的关系,可以实现更为准确的信息过滤。本课题尝试在上海交通大学网络学院已有的庞大在线学生群体和良好的网上学习条件之基础上,建立个性化课程学习平台,帮助、引导学习者进行有效的个性化学习。本文主要通过以下几点完成普适学习环境下的个性化服务:1)构建基于P2P网络的自组织社区为学习者提供不同的相似群体以便交流,解决学习者在分布式学习中的“孤独感”。2)通过自组织社区内同伴之间的相互推荐,解决“信息爆炸,但知识贫乏”中资源“大”的问题。3)增加教师端基于簇类的学习内容推荐,解决教师资源少不能对学习者进行有针对性指导的缺陷,并通过试题的定制让学习者更清晰地认识自身不足。4)根据不同学习者间学习历史记录和课程知识点的Ontology结构进行路径挖掘,解决学习者在“无序”的资源中缺乏有效引导而易重复摸索和“迷失”的现象,并对学习者的学习过程进行控制。基于角色的不同将实现的关键部分分为两个模块,分别为学习者参与的导学模块和教师参与的导教模块。其中导学模块中,通过自组织社区内的交流、资源推荐、学习路径挖掘(个性化课程编例)完成导学;在导教模块中,通过聚类分析,将具有相似特征的群体聚集在一起,完成教师对学生相对个性化的学习指导和各种资源的推荐。实验中,我们将该试验系统模拟应用于上海交通大学网络教育学院的《数据结构》课程,通过与传统学习平台的比较,以及与其它算法在课程中的成效比较,结果表明该种学习模式及相应的导学、导教方法大幅提升了学员的学习积极性和学习质量。同时,我们将导学方法进一步应用于网络课堂――《计算机应用基础》,得到了学员、老师的认可和一定的成效。在接下来的研究中,我们将不仅仅关注于用户的自组织、信息挖掘和推荐的准确度以及聚类特征的代表性,同时应该让学习平台更具智能性、更有交互性;并且能够根据环境和历史数据做出相应的推理和判断,适时地调整学习目标,并完成学习资源的自动推荐。更多还原

【Abstract】 With the rapid development and sweeping application of modern communication, the Internet and multi-media technology bring us with abundant learning resources and convenient, flexible E-Learning environment which breaks the physical limitation of traditional classroom-based learning mode. However, there are still a lot of new requirements and problems that should be solved in E-Learning environments, such as lack of learning control and personalized learning guidance. It becomes a big problem when an E-Learning system facing with a large number of learning resources and students with big gaps, and scant data of teachers’direct guidance. Besides, how to employ e-learners’feedback, such as the records of learning activity, testing and message, to generate personalized guidance content for the purpose of achieving the fixed learning goal of traversing all the Knowledge Points (concepts in a domain knowledge base) in a course, is also a big issue that attracts eyes of researchers on E-Learning.Recently domestic and foreign researches on such area have achieved a lot. With researches from the content-based filtering to collaborative filtering and then to self-organized community, and with the effort in creation of China E-Learning Technology Standardization, and with the introduction and application of Ontology, it is recognized that the information filtering should not only depend on the information itself, but also on the transmitters and users of the information. By analyzing the preferences and behaviors and making effective use of relationship of users, more accurate and efficient information filtering would come true.Based on the e-learners of a great number and the abundant learning resources in Network Education College of Shanghai Jiao Tong University, we construct a personalized learning platform to help and to guide the e-learner to achieve better learning quality. And the difficulties coming up are resolved in this way: 1) Establishing a self-organized community based on P2P to provide a different but similar cluster for different e-learners to settle the problem of e-learners’loneliness feeling in the distribute environment. 2) Adding recommendation functionality between partners in the same community when facing a large amount of learning resources. 3) Learning content recommendation based on clustering analysis that is performed by teachers for the lack of adaptive learning guidance from teachers. 4) Employing learning path mining methodology by analyzing historical data of learning records and ontology of the knowledge points for the disordered learning resources.Learning guidance module and teaching guidance module are the key parts of the system implemented. The main function of learning guidance module is to provide a communication platform for similar learners, including resource recommendation, learning path mining (sequencing), while the other one is to get learners with similar features around to make personalized guidance and recommendation from teachers easier.Fortunately, we got substantive support from the NEC of SJTU in experiment. By applying the methodology in“Data Structure”course, some comparison data between the new and old platform and between different algorithms are worked out. Experiments show that the model improves the quality of learning and enhances their learning initiative greatly. Meanwhile, we apply the learning guidance model to the real e-course of“Computer Application Basic”, and get a satisfying evaluation from both teachers and students.In the future research, we should not only focus on the accuracy of community self-organizing and recommendation, efficiency of information mining and representative of clustering features, but also concern about making the learning platform more intelligent and more interactive. That means, the platform will be able to reason and judge, and has great capability in adaptive learning and automatic recommendation according to the environment and historical data.更多还原