VR Presentation Training System Using Machine Learning Techniques for Automatic Evaluation

Abstract

In this paper, we build an immersive training space using building-scale VR, a technology that makes a virtual space based on an entire building existing in the real world. The space is used for presentations, allowing students to self-train. The results of a presentation are automatically evaluated by using machine learning or the like and fed back to the user. In this space, users can meet their past selves (more accurately, their avatars), so they can objectively observe their presentations and recognize weak points. We developed a mechanism for recording and reproducing activities in virtual space in detail and a mechanism for applying machine learning to activity records. With these mechanisms, a system for recording, reproducing, and automatically evaluating presentations was developed.