【作者】 朱郑州；

【导师】 吴中福；

【作者基本信息】 重庆大学 ， 计算机应用技术， 2008， 博士

【摘要】 随着网络技术的不断发展和教育技术的日益更新,现代远程教育(Modern Distance Education)的教育模式也正在发生改变,个性化、自主化以及协同学习(Cooperative Learning)等逐渐成为网络教育者和学习者追求的目标,如学习资源的自动提供,个性化学习方案的自动生成,学习服务(e-Learning Service)的自动发现,以及学习效果的自我评估。如何快速准确地发现教学过程中所需要的学习服务是影响教学效果的关键。传统的基于UDDI(Universal Description, Discovery, and Integration)的学习服务发现所采用的发现机制局限于关键字的匹配,是一种静态匹配的方式,尽管查找速度比较快,但自动化程度不高,而且不能保证找到所有满足需求的学习服务。由于本体(Ontology)具有共享、可重用等特点,有良好的概念层次结构及对逻辑推理的有效支持,且能从语义和知识的层次上描述信息系统的概念模型,成为语义网的重要技术之一。特别是基于OWL(Web Ontology Language)的本体技术可应用于网络教育,使得学习服务的描述具有语义信息,所以基于OWL-S(OWL for Service)的学习服务发现能够较好地克服UDDI匹配的弱点,提高学习服务发现的质量。然而该方法也还存在准确度低和效率低的问题,故论文结合本体论,应用二部图、粗糙集和用户满意度等理论,对学习服务发现算法进行了深入的研究。基于二部图(Bipartite Graph)的学习服务发现算法是把请求学习服务和发布学习服务的属性集分别作为二部图的顶点集,所有匹配属性之间的连线为边,边的权是属性匹配度,先把学习服务匹配问题转换为二部图的最优完全匹配问题,然后通过最优完全匹配问题的求解,实现学习服务的匹配,最终达到学习服务的发现。由于粗糙集理论(Rough Sets Theory,RST)可用于处理不精确、不一致、不完整的各种不完备信息,并从中发现隐含的知识,揭示潜在的规律,因此它特别适用于不要求精确数值结果的不确定性问题。基于RST的学习服务发现算法就是结合本体技术,把RST应用到学习服务发现当中。该算法是在学习服务匹配之前应用RST进行三步预处理操作:①规范化请求学习服务;②根据请求学习服务对发布学习服务进行不相关属性约减;③根据请求学习服务对发布学习服务进行依赖属性约减。其中的不相关属性约减和依赖属性约减可大大减少匹配的数量,从而提高学习服务发现的效率。然而,尽管采用了这些帮助提高学习服务发现查准率(Precision)、查全率(Recall)和效率(Efficiency)的算法,也只能尽快准确地查找到与学习者请求相匹配的学习服务,关键还在于学习服务本身的发现。在现有的学习服务发现系统中,根据学习者的请求能够找到一些学习服务,而且匹配度很高,但是学习者并不一定满意,这里的原因很多,其中一个主要原因就是现有的算法都只把学习服务自身的属性匹配度作为衡量学习服务匹配效果优劣的唯一指标,没有考虑到学习者的感受,这在一定程度上限制了学习服务发现系统性能的提高。因此本论文引入了用户满意度(User Satisfaction,US)的概念,提出了一种基于用户满意度的学习服务发现算法,该算法是把学习者对系统返回给他的学习服务的评价作为反馈信息,并设定一个修正函数,以动态更新发布学习服务的各个属性的匹配度权值,这不仅从客观上提高了学习服务的查准率,而且还从主观上提高了学习者对服务发现结果的满意程度。论文最后给出了一个CSCL原型系统,该系统集成了学习服务发现算法,学习者可以在该系统中进行协同学习,学习剧本包含有对所需学习服务信息的描述,学习者在学习过程中所需要的学习服务都可以由系统自动发现。更多还原

【Abstract】 With the development of network and educational technology, the educational pattern of modern distance education is changing gradually. Individualization, autonomous and coordination is becoming the goal of the tutor and the students under e-Learning environment. The variations of e-Learning educational goals include automatic push of learning resource, automatic generation of learning scheme and the discovery of learning software. So how to find the e-Learning service quickly and correctly is becoming the key point that influences education effect.The traditional e-Learning service discovery mechanism is based on UDDI. It is limited in keywords matchmaking, and its matchmaking mode is static. Although its searching rate is rather rapid. It cannot ensure finding the required e-Learning service which satisfies the student’s needs very correctly, and its automation degree is not very high.Because Ontology owns the characteristic of share and reuse, and it has good concept structure, and it support logic inference. From 1990s, Ontology was applied on many fields such as knowledge engineering, information retrieval, information integration and knowledge management, and was become one of the important technologies in semantic web.Ontology applies on e-Learning service description, making service description information own the semantic information. The e-Learning service discovery model based on OWL-S can overcome some of the weakness of the discovery model based on UDDI, and can improve the precision and recall of e-Learning service discovery. But this method still has the some problems: its precision is not very high and efficiency is low. According to ontology, bipartite graph, rough sets theory and user satisfaction, the dissertation proposed 3 algorithms.The first algorithm is e-Learning service discovery algorithm based on bipartite graph which called eLSDA-BG. In this algorithm, the property sets of the required e-Learning service and the advertised e-Learning service are becoming vertex sets of a bipartite graph. The edges of this bipartite graph are the line between the matchmaking properties. The weight of the edges is property matchmaking degree. So the problem of e-Learning service matchmaking becomes the optimal complete matching of bipartite graph. Because Rough Sets Theory can find implicit knowledge, reveal the law from inaccuracy inconsistency imperfect information. I applied rough sets theory on e-Learning service discovery, and developed an e-Learning service discovery algorithm named eLSDA-RS. Algorithm eLSDA-RS combined Ontology technology and rough sets theory. There are 3 operations before e-Learning service matching. The first is to normalize the required e-Learning service. The second is to reduce uncorrelated attributes of advertised e-Learning service according to required e-Learning service. The third is to reduce dependency attributes of advertised e-Learning service according to required e-Learning service. The last 2 steps can decrease the number of advertise e-Learning service which should be match made, and improve the efficiency of e-Learning service discovery.Although many algorithms can improve the precision, recall and efficiency of e-Learning service discovery. The students are uncertain satisfied with the results. One of the reasons is that all of the algorithms regard properties matchmaking as the only indicatrix without considering the students’ feelings. So I led-in a new factor -- User Satisfaction which is the user’s feeling to the result of e-Learning service discovery. And I proposed a new e-Learning service discovery algorithm based user satisfaction called eLSDA-US. This algorithm allows the students to take part in the process of e-Learning service discovery, and also allows them to evaluate the result of service discovery. The students’ evaluation in the form of User Satisfaction is fed back to the system. Adopting an amendatory function which takes the User Satisfaction as input, the system modifies the weights of each property of the advertise service, and then the total match degree of service discovery will up to best. I adopt two methods to encourage users to use the e-Learning service discovery system.At the end of the dissertation, a prototype system of CSCL is proposed.更多还原