基于交互行为感知的MOOC学习导引机制研究

To mitigate the lack of the appropriate interventions on the Massive Online Open Courses(MOOC) caused by the high learner-teacher ratio and the high degree of freedom, on the basis of the measurement and analysis on the behaviors of the learners, this research will attempt to build some automatical learning guiding mechanisms on both macroscopic and microcosmic aspects. On the macroscopic aspect: The learning communities of interests will be constructed by importing overlapping communities detecting algorithms, through which learners can be organized to engage in the cooperative learning activities such as Question-and-Answer cooperatively; How to arrange the learning sequence of courses will also be considered in this research. By constructing an appropriate topology of courses, learners can be navigated to reasonable learning paths. On the microcosmic aspect: Proper learning activities and key points of courses can be labeled by mining the learning activity patterns of the learners, thus hints can be sent to learners to correct their learning habits; Knowledge from multiple sources will also be aggregated into a knowledge graph, and then integrated in the lecture learning processes for avoiding switching between lecture learning and information searching, thus learners can consentrate on how to learn on pre-defined tracks. At last, the guiding mechanisms will be evaluated by large scale simulations and tested by deploying in real applications, which will provide theoretical and practical basis on how to interpose the learning processes correctly to improve the learning effects.

针对MOOC的高师生比和高自由度带来的学习过程干预不足的问题，在对学习者行为进行测量分析的基础上，从宏观和微观两个层面建立自动学习导引机制。宏观层面上，利用重叠社区发现机制，建立基于兴趣相似度的学习共同体，组织学习者开展如互助问答等协同学习活动；挖掘课程间的继起关系，构建合理的课程知识拓扑，给予学习者恰当的学习路径导引。微观层面上，通过对优质学习者行为模式的挖掘，标注正确学习行为和关键知识点，对学习者行为进行提示，矫正用户学习习惯；聚合多个知识源构建知识图谱并融入视频学习过程，避免频繁在视频学习与资料查阅间切换思路，专注投入预定学习轨道。最后，通过大规模仿真和真实环境下的部署测试，对导引机制的效能进行评估，为正确干预学习过程，提高学习效果提供理论和实践依据。

针对MOOC的高师生比和高自由度带来的学习过程干预不足的问题，在对学习者行为进行测量分析的基础上，从宏观和微观两个层面建立自动学习导引机制。宏观层面上，爬取主流MOOC网站的评论区数据，利用复杂网络社区发现机制，建立了基于兴趣相似度的学习共同体，并提出了一种可对学习者在网络中的程度进行排序的算法；微观层面上，基于真实课堂调查数据，分析了不同风格学习者知识整合的差异性，并组织学习者开展同伴互评等互助学习活动，提出了支持恶意评价识别的互评方法；最后，聚合多知识源构建知识图谱并融入视频学习过程，避免频繁在视频学习与资料查阅间切换思路，专注投入预定学习轨道，并以浏览器插件的形式进行呈现，为正确干预学习过程，提高学习效果提供理论和实践依据。