Enhanced Knowledge Graph for True-false Question Generation: A Case Study in Computer Networks Course

Knowledge graph provides an ideal technical means to automatic question generation. The classical knowledge graph contains direct relationship between entities and their attributes, but it is difficult to contain implicit logical relationships, like evolution, analogy and causality. This is not conducive to the generation for logical questions. In this paper, we analyzed true-false question features in computer networks course and classified questions into three types: fact discrimination(FD), conception discrimination(CD) and logic discrimination(LD). In order to generate these three types of question, we propose an enhanced knowledge graph (E-KG). By upgrading model layer and data layer, we can use inheritance relationship to describe knowledge points from the same knowledge domain. This helps to find analogical knowledge more conveniently. We label the implicit logical relationships between entities and use reference sentence to support and explain these logical relationships. Finally, we construct an E-KG by using computer networks knowledge and analysis the true-false question generated by our E-KG to verify the help of E-KG with question generation.