Automated Prediction of Item Difficulty in Reading Comprehension Using Long Short-Term Memory

Abstract

Standardized tests are an important tool in education. During the test preparation process, the difficulty of each test item needs to be defined, which previously relied on expert validation or pretest for the most part, requiring a considerable amount of labor and cost. These problems can be overcome by using machines to predict the difficulty of the test items. In this study, long short-term memory (LSTM) will be used to predict the test item difficulty in reading comprehension. Experimental results show that the proposed method has a good prediction for agreement rate.