Teacher-student dialog is one of the important components of classroom teaching, which is not only a tool for transferring knowledge, but also a reflection of the transformation of teacher-student role relationships, teaching methods, students’ learning status, and teaching effectiveness. Therefore, the identification and statistics of teacher-student dialogues have once become the main entry point for the application of artificial intelligence technology in classroom teaching analysis. In this study, through careful observation and in-depth analysis of the speech characteristics of teachers and students in classroom interactions, we explore and understand the multiple roles played by teachers and students in the educational process, and construct the intelligent localization model of teacher-student roles in BERT-CNN-BiLSTM, which basically realizes the automatic localization of the roles of teachers and students in the classroom teaching, brings new perspectives and tools to the research and practice of the classroom teaching analysis field and assists in the teacher's professional development.