联合知识图谱和时间特性的数学知识自动推荐方法

Automatic Recommendation Method of Mathematical Knowledge Based on Combined Knowledge Graph and Temporal Feature

在教育领域的自动推荐方法大多通过历史内容的相似性进行推荐,未同时考虑用户交互序列和知识与知识之间的内在关联信息,难以精准地向学习者推荐其所需知识点。提出一种基于联合知识图谱和时间特性的数学知识自动推荐方法。该方法首先利用自注意力机制和前馈神经网络获取带有时间特性的学习者表示,然后根据知识图谱三元组中知识点与知识点的高阶连通特性和学习者特性深层次表征知识点,最后计算学习者与知识点交互的概率,并根据概率进行推荐。在自构建的初中数学知识推荐数据集上测试表明,本文方法比几类经典的基准系统在AUC值、精确率、召回率和F值上都有提升。该方法的提出能够更加准确地为学习者推荐知识点,帮助其构建知识体系,可望为学习者的自适应性和个性化学习提供了支持。

The automatic recommendation methods in the education field mostly use the similarity of historical content to recom-mend, and do not consider the user interaction sequence and the internal correlation information between knowledge and knowledge at the same time, so it is difficult to accurately recommend the knowledge points they need to learners. Therefore, the paper proposes an automatic recommendation method for mathematical knowledge combining knowledge graph and time characteristics.The method first uses self-attention mechanism and feed-forward neural network to obtain learner representation with time characteristics, and the knowledge points are represented in depth according to the high-level connectivity between the knowledge points in the knowledge graph triples charateristics of learners. Finally, the probability of the interaction between the learner and the knowledge points is calculated, and recommendations are made based on the probability. Tests on the self-built junior high school mathematics knowledge recommendation data set show that the method proposed in the paper has improved at AUC,precision,recall and F-scores respectively, compared with several types of classic benchmark systems. The proposed method can more accurately recommend knowledge points to learners’ and help to build a knowledge system, and provide certain support for learners’ adaptive and personalized learning.