[1M5-OS-14b] OS-14

[1M5-OS-14b-02] A study of element and relationship classification methods using BERT for argumentation structure visualization systems

〇Satoru Kogure¹, Sora Toyama¹, Yasuhiro Noguchi¹, Koichi Yamashita², Tatsuhiro Konishi¹ (1. Shizuoka University, 2. Tokoha University)

Keywords:Argument structure visualization, Machine learning, Element classification, Relationship classification

When teachers evaluate student reports at universities and other institutions, it is necessary to check whether the students’ own arguments are correctly explained based on evidence such as experimental results and other obvious assumptions. However, it is difficult for a single teacher to evaluate the argumentation structure of all class reports. In response to these problems, we have developed an argumentation structure visualization system. This system can be used by students to examine the argumentation structure of their own reports before submitting them, and it can also be used by teachers to check the argumentation structure of submitted student reports. We propose a method using BERT and other methods for ”element classification” and ”relation classification” of sentences, which are necessary for this system to visualize argumentation structure.