Examining Rating Scales for General Speaking Proficiency and Reconstructing a Database of Teachers' Ratings of Speech Samples

本研究首先调查了中学和大学英语教师对学生英语演讲和互动录音的评分。为了明确数据集的特征，采用概化理论、项目反应理论和潜在等级理论对评分进行分析。其次，采用自组织图（SOM）的数据挖掘方法对同一数据集进行分析，并将其特征可视化成二维图；最后，研究阐明了如何从数据挖掘方法获得的更简单的图中推断出高度统计分析的结果。

This study first investigated ratings which English teachers in secondary and tertiary education had given of recordings of students' speeches and interactions in English. In order to clarify the characteristics of the dataset, the ratings were analyzed using Generalizability Theory, Item Response Theory and Latent Rank Theory. Secondly, the same dataset was analyzed by a data mining method called Self-Organizing Map (SOM), and its characteristics were visualized in a two-dimensional map. Finally, the study clarified how the results of the highly statistical analyses can be inferred from a simpler map obtained by the data mining method.