The purpose of this study was to examine a new method of classroom speech analysis through the analysis of classroom speech using the Large Language Model (LLM). As a method of the study, an elementary school in-service teacher's math class was recorded, and the transcribed speech protocol was used for labeling from the viewpoint of “independent, interactive, and deep learning. After LLM learned the labeled speech protocols and constructed an AI model, we conducted a speech analysis on “proactive, interactive, and deep learning. As a result, it was inferred that the accuracy of AI's judgment of “proactive learning” tends to be relatively high, while “interactive learning” is more likely to be judged by words such as “exchange,” “others,” “consultation,” and so on. Since many subject-specific words are used for “deep learning,” it is necessary to further increase the amount of data accumulated to improve the accuracy of the judgments.
