人工智能赋能个性化学习 : E-Learning 推荐系统研究热点与展望

E-Learning领域的推荐系统在满足学习者个性化学习需求方面发挥着重要作用。近年来，国际上围绕E-Learning 推荐系统开展的研究迅速增多。采用文献计量分析方法对该领域的研究进行系统分析，有助于为E-Learning推荐系统的高水平研究和高质量应用提供镜鉴。综括而言，当前国际E-Learning领域的推荐系统研究热点及其演变趋势集中体现在6个方面：一是融合多种技术优势的混合推荐日益受到重视且逐渐成为主流。二是伴随技术支持下群体学习的多元发展，个性化推荐由关注个体推荐逐步转向关注群体推荐。三是随着大规模开放在线课程的流行，个性化推荐逐步突破小规模而面向大规模学习者群体，重视通过对海量学习资源和过程数据的搜集和挖掘而提供个性化推荐。四是从心理学层面关注学习者情绪变化，并据此构建上下文推荐系统，通过优化调整推荐内容不断促进学习者高效完成学习任务。五是在推荐功能上更加强调学习模型构建，重视提升学习者的深层次认知能力和促进有效学习。六是在先进技术的支持上，个性化推荐系统强调引入深度学习技术，不断优化其表征能力、融合效率和推荐效果。

Recommender systems in the field of E-learning are essential to meeting learners’ personalized learning needs. In recent years, research on E-learning recommender systems has grown rapidly worldwide. To systematically analyze relevant research in this field with the method of bibliometric analysis helps provide a reference for the high-level research and high-quality application concerning recommender systems in E-learning. The analysis results show the research hotspots and their evolution tendencies in this field from six aspects. First, the hybrid recommender system that takes advantage of multiple recommendation techniques has received increasing attention and has developed into the mainstream technique. Second, with the diversified development of group learning supported by innovative technologies, the focus of personalized recommendation has shifted from individual recommendation to group recommendation. Third, with the popularity of massive open online courses, there is a trend toward personalized recommendation on a large scale through collecting and mining large volume of data related to learning processes and materials. Fourth, attention has been paid to the emotional changes of learners from a psychological perspective, based on which context-aware recommender systems can be constructed to constantly promote learners to learn efficiently through adjustment and optimization of the recommended learning content. Fifth, increasing emphasis has been placed on learning model construction, with a focus on improving higher-order cognitive skills and promoting effective learning. Furthermore, supported by advanced deep learning technologies, the representation ability, information fusion efficiency, and recommendation effectiveness of personalized recommender systems can be continuously improved.