基于知识图谱的微信智能刷题系统

Intelligent WeChat Exercise System Based on Knowledge Graph

针对各类做题软件中做题模式单一导致学生复习效率低下的问题,将传统的协同过滤算法与知识图谱进行融合,构建了一套微信智能刷题系统。在知识图谱中构建出用户、题目、知识点三类节点之间的关系,将用户、题目之间的相似度作为图的权重。考虑到未掌握知识点是学生答错题的主要原因,采用基于知识的方法作为主要推荐方式,从知识图谱中查询出错题所涉及的知识点加入队列等待推荐。根据用户历史的刷题数据,每天定时将相似用户的错题进行推荐。近2年的实验结果表明,该模式受到了学生的广泛青睐,可以有效地提升学生的复习效率。

Focusing on the problem of students’ low efficiency in tradition exercise systems which are with a similar mode, an intelligent exercise system based on WeChat platform was developed. The system combines traditional collaborative filtering recommendation algorithm and knowledge graph. First, the relationship among user, question and knowledge point in the knowledge graph is considered, and the similarity between question and user’s answer is used to yield weights. Considering that the related knowledge point having not acquired is the main reason of getting the wrong answer, a knowledge-based recommendation scheme is used as the main method. Related knowledge points and similar questions queried from knowledge graph will be pushed into to queue and wait for recommendation. Furthermore, according to the historical exercise data, the wrong questions of the similar users will be recommended everyday as well. It was used for two years, and experimental results show that this mode has received wide favor among students and it can raise the review efficiency.