普适计算技术推动下的泛在学习已经成为当前教育技术领域的研究热点和下一代e-Learning的重要发展方向。无处不在的学习需要无处不在的海量的学习资源，需要能满足各种个性化需求的、可进化的、情境化的、适应性的学习资源。可进化性已经成为未来学习资源的重要特征和趋势，将对当前更新缓慢、缺乏进化动力的e-Learning资源带来巨大的挑战。Web 2.0理念的快速传播及其成功实践的发展表明，充分利用和聚合普通用户的群体智慧比Web1.0时代的单向信息传递模式更具活力。世界上并不存在一个上帝或超人可以生产所有的资源，泛在学习必然采取“群建共享”的开放协同建设模式，来实现泛在学习资源的持续生成和进化，满足学习者不断增长的各种个性化需求。然而，“群建共享”的开放协同建设模式也存在一大缺陷。开放的环境中用户群体复杂，任何人都可以参与资源建设，将直接导致资源质量良莠不齐，进而引起维基百科所遇到的“信任危机”。究其原因，完全开放的环境下学习资源处于“无监管”状态，缺乏对其进化方向的控制和约束，导致资源毫无方向的“杂乱”生长。此外，当前的e-Learning资源多采用静态元数据的描述技术，资源之间普遍缺乏语义关联，不利于学习者的知识导航和有效学习。常常采用人工编辑超链接的方式（如维基词条间的关联）建立资源间的关联，耗时耗力，难以快速发现、动态建立资源间的语义关系。本研究基于生态学视角将学习资源视为泛在学习生态中的有机生命体，从内容进化和关联进化两个方面研究了资源进化的模式，重点解决了影响泛在学习环境下开放学习资源有序进化的两个关键问题：一是学习资源内容进化中的智能有序控制问题，即如何实现普通用户进行内容编辑时的智能审核，什么人在什么时间编辑的什么内容可以被系统自动接受或拒绝；二是学习资源关联进化中的动态语义关联与聚合问题，即资源实体之间如何动态建立语义关联，并在资源语义关联信息的基础上自动聚合生成更大粒度的资源群。本研究开展了如下三个方面的工作：首先，本研究基于语义基因和信任评估模型探索出一种可以有效实现对资源内容进化过程智能控制的方法。核心思想是依据编辑内容与资源语义基因的语义相似度和用户的信任级别两方面的信息综合判断每次内容编辑的可信程度，最终决定是否采纳此次内容编辑，从而达到对内容进化过程的智能控制。检验结果显示，该方法可以有效控制资源内容的进化，实现对内容版本更迭的智能审核，大大减轻资源创建者人工审核内容的负担，能更好、更快的促进学习资源内容的持续进化。其次，本研究综合应用语义基因、规则推理、关联规则挖掘等技术提出一种可以有效实现资源动态语义关联与聚合的解决方案，可以弥补传统人工资源关联和聚合时耗时、耗力、主观性强等方面的缺陷。动态语义关联技术可以在资源空间的结点间建立起丰富的语义关系，形成可持续扩展的资源语义关系网络。动态语义聚合技术可以从大量的关系中挖掘出更大粒度的、有意义的资源结构体（主题资源圈、有序知识链）。检验结果显示，本研究提出的动态语义关联与聚合方法具有较高的准确性，可以满足开放环境下资源关联进化的需要。最后，本研究依托学习元平台（Learning Cell Knowledge Community System, LCS），综合应用J2EE、本体、信任评估模型、基于规则推理、关联规则挖掘、Flex等技术开发了泛在学习资源进化支撑系统，核心包括开放内容编辑、知识本体管理、内容进化控制、动态语义关联、动态语义聚合和可视化进化路径展现等功能模块，用于支持LCS中的学习资源进化。此外，本研究应用案例分析法和问卷调查法对LCS中的资源进化过程和进化效果进行了分析评价。通过上述工作的开展，很好地解决了本研究提出的影响资源有序进化的两大关键问题（学习资源内容进化中的智能有序控制问题、学习资源关联进化中的动态语义关联与聚合问题），并在如下两方面取得了一定的创新：(一) 基于生态学视角，从生命有机体的角度看待泛在学习资源，聚焦学习资源有序进化，推动泛在学习资源层面的研究。生态学中“开放、动态、关联、进化”的观点和原则，对于开展泛在学习资源研究具有重要指导意义。本研究基于生态学视角，从生命有机体的角度看待学习资源，重点对学习资源的有序进化问题进行探索性研究，一定程度上弥补了当前泛在学习在资源进化方面研究的不足，推动泛在学习资源层面的发展。(二) 针对学习资源内容进化中的智能有序控制和关联进化中的动态语义关联与聚合问题，提出一套有效的工程化解决方法，实现开放环境下学习资源的有序进化。本研究综合应用语义Web、信任评估、规则推理、关联规则挖掘、可视化等技术提出一套可以实现学习资源内容进化中的智能有序控制、关联进化中的动态语义关联与聚合的解决方法，并依托LCS设计开发了资源有序进化的支撑环境，验证了该方法的有效性。本研究的开展能够更好的促进高质量开放资源的协同建设、传播和共享，对于泛在学习资源的生产与管理，不仅具有理论上的指导意义，而且具有极大的实践价值，可以采用工程化的方式推广应用到其它开放资源建设环境中。

Driven by ubiquitous computing technology, ubiquitous learning (u-Learning) has become the future way of e-learning. U-Learning needs ubiquitous and large amount of learning resources with characteristics of personalization, evolvability, contextualization and adaptability. Evolvability has become a key feature and trend of future u-Learning resource, which will give a big challenge to current e-Learning resources with less dynamics of evolution. The rapid spread of Web 2.0 concept and its successful practices fully proves that the aggregation of mass wisdom can make information generation and transfer more vigorous than that in Web 1.0 era. As we all know, there is no “God” or “Superman” who can generate all learning resources for u-Learning. For this reason, we must adopt an open and collaborative method to construct u-Learning resources.However, there is an inevitable problem while using the open and collaborative resource construction method. In open environments, everyone can engage in constructing learning resources, and users have different discipline backgrounds and expertise, which will lead to trust crisis encountered by Wikipedia. The reason for that is the discursive growth and in poor control of quality of learning resource. In addition, we always use static metadata technology to describe learning resources with weak semantics. Semantic associations between resources are obviously inadequate and mainly built by hand, which cost lots of time and manpower, and are not good for knowledge navigating and effective learning.From the ecological perspective, this research considers learning resources as lively entities in u-Learning system, and studies the resource evolution models from two aspects of content evolution and association evolution. This research focuses on solving two key issues influencing resources’ orderly evolution. One is how to control content evolution intelligently. The other is how to implement semantic associations and aggregations dynamically. In this research, three aspects are mainly studied as follows:Firstly, a method of controlling content evolution processes intelligently using semantic gene technology and trust evaluation technology is proposed. An editorial credible degree is an important indicator deciding the result of content revision every time. It is counted according to the semantic similarity between new content and current resource’s semantic gene and user’s trust level synthetically. Experimental results show that this method can automatically audit content versions and control content evolution effectively. It can reduce artificial burden greatly and gear up resource content’s continual evolution.Secondly, this research puts forward a kind of solution to make semantic association and aggregation dynamically using semantic gene, rule based reasoning and association rule mining technologies comprehensibly. Through the application of this solution, rich and semantic associations can be built in resource cyber space, and more significant and lager grained resource group can be found. Experimental results show that this method has a high accuracy and can support resource association evolution effectively.Finally, this research develops a u-Leanring resource evolution supporting system based on Learning Cell System(LCS) through J2EE, Ontology, Trust evaluation, Rule based reasoning, Associate rules mining, Flex, etc. This system consists of six functional modules, including open content editing, knowledge ontology management, content evolution control, dynamic and semantic association, dynamic and semantic aggregation, and visual evolution path presentation. In addition, this research applies case analysis and questionnaire to evaluate the effect of resource evolution in LCS.This research has contributions to the u-Learning research field and achieves two innovations as follows: (1) From the ecological perspetive, considering learning resource as an lively entity, this research focus on studying u-learning resource evolution. To some extent, this research makes up the shortage of u-Learning resource, especially on the aspect of resource evolvability, which will promote the development of u-Leanring research. (2) This research adopts semantic web, trust evaluation, rule based reasoning, association rule mining and visulization technologies innovatively to solve two key issues influencing resources’ orderly evolution. Based on LCS, a system of supporting learning resource evolution is developed. The achievements of this research has a great potential for promoting the construction, spread, and share of open learning resources with high quality.