互联网的发展和移动终端的普及使得学习的发生变得情境化、个性化，个性化学习资源成为满足个体学习需求的重要条件。而在学习过程中，学习者的情境、偏好、学习风格、认知能力、典型问题等均处于动态变化之中，这种动态变化不仅影响学习者对内容的需求，也对内容的组织和呈现提出要求，这就要求基于上述特征并从教学与学习设计角度对资源进行设计，即设计过程除了实现内容的适应，还需要从教学与学习设计角度根据学习者的情境和个体特征提供适应性的组织结构。目前在线学习中学习资源的适应主要基于推荐和聚合技术，它们通过标签、兴趣或主题相似度算法实现了学习者需求与资源库中已有资源的匹配。两种方式推荐的学习资源多是针对特定学习需求预设的，其结构和组织形式静态不变，不能适应学习情境和需求的动态变化。而推荐的资源仅将学习者特征作为内容推荐的指标，缺少对资源组织结构的设计，因此在支持适应性学习方面存在局限。为满足学习者动态变化、结构化的资源需求，要改变传统预设型的资源推荐方式，寻求一种以教与学理论为指导并依据个体特征进行动态适应的资源设计和生成方式。

为实现这一目的，本研究基于教学与学习设计理论，参照资源推荐、聚合以及其他领域内容生成方法，通过分析个性化学习过程的需求及资源特征，提出了一种个性化学习资源自动生成方法，该方法区别于传统的资源推荐和聚合，它基于对个性化学习资源需求的分析，一方面构建学习资源结构化表征模型，定义了个性化学习资源的基本要素及其动态结构，使得学习资源的内容与结构分离，为个性化学习资源的动态适应性组织提供基础；另一方面，基于教学与学习设计的流程，设计了个性化学习资源自动生成机制，界定学习资源生成的基本流程，即首先基于学习主题对相应的素材进行汇聚形成个性化学习资源聚合容器（Container）；其次基于情境感知和计算技术对学习者的情境、风格、知识结构、能力、典型问题等进行计算并确定其学习目标；最后基于获取的特征和资源生成算法进行素材提取和结构化组织，形成支持个性化学习的资源，实现资源设计的自动化。

基于提出的学习资源结构化表征模型和自动生成机制，研究者设计并开发了个性化学习资源自动生成系统，将其应用到中国厦门H高校为期4周的留学生课程——【发展汉语课程听写典型问题】，用于支持学习者泛在条件下的个性化学习。实验结果表明，在系统生成准确性评价方面，系统在满足学习者需求方面具备较高的准确性，同时让学习者具有较好的感知适应性，保证了学习者较高的态度、科技接受度和满意度。在促进学习者个体学习方面，由于系统可以基于个体的情境和特征对实验组的资源进行动态设计和组织，实验组学习者可以在不同情境下习得与情境相关的内容，其对学习内容的理解更加全面，参与的活动更丰富，也使得其学习时长和投入更多，最终促进了其学业成绩的提升；在认知负荷方面，实验组在高认知投入的前提下，其认知负荷与对照组没有显著差异，一定程度也表明系统在促进个性化学习过程中发挥了积极作用；学习者和教师的访谈结果也表明，系统对促进学习的个性化、降低教师内容设计难度方面也具有积极作用。该课程结束后，研究者再次将系统用于北京某学校的【小学科学光合作用】课程验证其迁移效果，实验结果表明系统生成资源的准确性较高，实验组学习者在认知负荷未明显增加的前提下，学业成绩上同样显著优于对照组。

从两个研究的结果可以得出结论，本研究提出的个性化学习资源自动生成系统可以有效为学习者提供适应性良好、结构化的优质资源，可以有效促进学习者个性化学习的效果，与此同时，系统具有良好的场景可迁移性，具备实践应用的潜力。

本研究的创新点聚焦于当前资源供给方式忽视资源结构设计的问题，通过分析资源需求，建立个性化学习资源结构化表征模型，实现内容与结构的分离，从而更好地支持资源的动态变化；同时基于教学与学习设计规则设计了个性化学习资源自动生成机制，通过对素材的汇聚、学习者需求的计算以及素材的动态组织，实现了个性化学习资源的自动生成，支持了在线学习中动态、结构化的资源供给。

未来的研究中，一方面将聚焦于个性化学习资源内部要素的有序组织，探索对各要素进行科学排列的方法，使资源内容的组织能够更好支持个性化学习；另一方面结合脑电技术、眼动技术，揭示系统在影响学习者认知和行为方面的内在原因，进而确定设计机制的科学性，以辅助个性化学习资源自动生成机制的改进。

The development of internet and mobile technologies make the learning process more contextualized and personalized, therefore, personalized learning resources become one of the most important factors to meet learners’ personalized requirement. During this process, learners’ context, preference, learning style, cognitive capacity and problems are dynamically changing, while these changes not only influence learners’ content needs but also the content organization. Thus, there is a need to design learning resources based on the information so as to come up with well-organized resources for learners’ personalized learning. To achieve this target, resources should first be adapted on the content and then organized based on the theory of instructional and learning design. Currently, technologies for the adaption of online learning resources are recommendation and aggregation. They mainly realize the adaption by matching the similarity between learners and resources based on specific tags or characteristics. Resources provided by these methods are usually for predesigned learning needs, while the content and organization could not dynamically adapt to the change of the learning context. Moreover, as there lacks design for resources’ structure, the methods could be limited in supporting learners’ learning. In order to provide learners with structured resource which could dynamically adapt to the context, it is necessary to change the traditional resource provision methods and investigate a new method guided by instructional and learning design theories.

To solve the abovementioned problem, this research comes up with a personalized learning resource generation method by analyzing the requirements of personalized learning and referencing related resource provision methods such as recommendation, aggregation, and content generation. The proposed method is different from recommendation and aggregation on two aspects. Firstly, it proposes a structure model for personalized learning resource. In the model, factors and their organization are defined in detail by analyzing the need of personalized learning. This makes it possible to separate the learning content from the structure which make the learning resources more flexible. Secondly, the method designs the mechanism and processes for the generation of learning resources which involves the aggregation of the Container, the computing of the personalized information and learning target, and the generation and organization of personalized learning resource based on specific algorithm and the materials in the Container. This method makes it possible to automate instructors’ resource design work.

Based on the proposed resource model and generation mechanism, the researcher developed the automatic generation system for personalized learning resources and used it in a 4-week course (Typical Problems for Chinese Dictation) for foreign under-graduates in Xiamen, China, to support their personalized learning. From the perspective of resource quality generated by the system, result suggests that the generated resources are of high-quality and learners have better sense on the adaption of the provided resources because the system has designed the resources according to learners’ needs. Moreover, the system guarantees learners’ high attitude, technology acceptance and satisfactory during the learning process. From the perspective of promoting learners’ personalized learning, the system could provide experiment group learners with resources related to different context based on learners’ characteristics. Under this circumstance, learners participated in more activities and had deeper learning engagement. As a result, they had more comprehensive understanding for specific knowledge and achieved better performance. What’s more, experiment group did not show significant differences on cognitive load compared with control group, even though they have deeper engagement which verified the positive effect of the system. At the end of the experiment, interviews on learners and instructor showed that the system could promote learners’ personalized learning and reduce instructor’s resource design difficulty. After this experiment, the system was then used in another course (photosynthesis) in a primary school in Beijing, in order to see if the system and mechanism could migrate among different contexts. Results showed that the resource generated by the system were also of high quality and learners in experiment group outperformed control group while they did not show higher cognitive load.

In conclusion, these experiments demonstrated that the proposed personalized learning resource generation system could provide learners with adaptive learning resources which were well organized. Also it could promote learners’ learning performance. These results reveal that the proposed method and system in this research work well in supporting personalized learning under different contexts and have potential for large scale applications.

Innovations of this study focus on the problem that there lacks structure design for online learning resources in current applications and comes up with a new method to support personalized learning. The method established a learning resource structure model based on personalized learning requirements and realized the separation of resources’ content and structure. On the other hand, this research designed the mechanism that support resource generation based on instructional and learning design theory. This makes it possible to aggregate, organize materials and then generate adaptive resources dynamically based on learners’ characteristics, finally realized the dynamic, structured resource provision.

In the future, the researcher will focus on well-ordered organization of factors in resources and investigate new method which support more scientific organization of the extracted materials. What’s more, in order to reveal the deep level mechanism why the system affect learners’ cognition and behavior, the researcher will investigate EEG, eye-tracking techniques and improve the mechanism based on the multi-modal result.