Comment Evaluation by Combining Comment and Word Mutual Evaluation Method and LSTM Evaluation Method in Lecture Questionnaire

Many universities give free-description question-naires to students to obtain feedback on faculty development (FD). When this is done, a proper analysis of the students’ comments is necessary. The number of comments from the free description that can be acquired for the FD activities is often not very large. To evaluate a small amount of data with approximately 1500 comments here needs to be some improvement in the currently available evaluation methods. In this study, we propose a probability distribution for the evaluation. We also propose a method for mutually evaluating the words and the comments long with the LSTM evaluation method by using neural networks. However, these methods seem to have differences in accuracies between the estimated values of the closed tests and the unrated comments. Therefore, we apply two methods to the bootstrap method to estimate the unrated comments; we also propose a method to incorporate the comments into our solution.