Student Questionnaire Analyses Using the Clustering Method based on the PLSI Model

This paper proposes a novel document clustering method based on Probabilistic Latent Semantic Indexing (PLSI) model. Recently, it is more strongly expected to improve quality of the education in universities. Student questionnaire is an effective way from which we can know the student’s evaluation and make use it to improve the education quality. Free text format answer can be used to obtain in particular the opinion of the students directly. On the other hand, in the field of natural language processing, the various techniques and models for document classification or document clustering problems are developed. By the classification or the clustering for the questionnaire document, it is possible to make groups of students with the similar opinion. When a document set is large-scale, it is shown that good classification accuracy can be achieved in many studies. For a comparatively small-scale document set, the method using the PLSI model studied in recent years is remarkable, and it can achieve good classification accuracy to such small size of documents. In this study, our target documents for analysis are the student questionnaire with text free format for classrooms in university. Then, it is premised on the following. (1) The document set is comparatively small-scale. (2) The category which should be classified is not clear in advance. By the clustering method using the PLSI model, the maximum likelihood estimator of the parameters of assume