Presentation Information
[5Yin-A-35]A study on generating explanations for diagrams and tables to predict correct answer rates in mock exams
〇KAORU HIRAMATSU1 (1. Saitama University)
Keywords:
academic achievement test,Generative AI,correct answer rate
When creating test questions, it is necessary to have an appropriate variation in the difficulty of the questions so that the variation in learners' understanding is close to the score distribution, and it is thought that highly accurate prediction of the correct answer rate will be one of the useful indicators when creating questions. Therefore, we report the results of introducing features extracted from the explanations of diagrams using generative AI into a model that predicts the correct answer rate from the features of mock test questions that we have developed so far, and verifying the effect of this on improving the accuracy of correct answer rate prediction.