This paper described a presentation robot system that interacts with participants in a classroom. Especially, we have proposed the robot interaction to interrupt the main scenario and speak short comment based on a state of participants. For the robot's presentation, it is important to involve the participants. We are developing a presentation robot system to give a lecture for a large group at the university. Especially, we focus on the interrupt behavior of a human being based on a state of participants to involve participants. Thus, we propose the measurement of the state of participants based on face direction detection. We have developed the integrated system of a humanoid robot PALRO named Masaru, a data management server and client PCs for each participant. For detecting face direction, we apply AlexNet to divide a face direction into 4 categories corresponding with the motivation for a presentation. As an experiment, we performed a demonstration in a lecture of robotics by 55 participants. MASARU interrupts the main scenario and takes action based on the face direction result. MASARU's presentation is mostly met with acceptance shown by the evaluation button of an application and the questionnaire.