Proceedings of the Annual Conference of JSAI
Online ISSN : 2758-7347
37th (2023)
Session ID : 4R2-OS-22a-02
Conference information

Evaluation Method for Debate with text-vocal-video information
*Motoki SAKAIMasaki SHUZO
Author information
CONFERENCE PROCEEDINGS FREE ACCESS

Details
Abstract

Many companies adopt group discussion (GD) in recruitment selection, students have to train their communication skills to receive unofficial job offer. Therefore, this research proposes a short time debate-based training system to improve communication skills. In experiment, 43 short time debates were conducted, and movies, audios, speech transcriptions data, and impression and comprehension labels were collected. In this study, only speech transcriptions data were used to estimate listener impression toward speaker utterances. For Estimation of listener impression, the pre-trained BERT released by Tohoku University was adopted. As a result, the proposed method could not accurately estimate listener impressions. In a future work, we are going to improve an estimation method in addition to other information such as movies and audios.

Content from these authors
© 2023 The Japanese Society for Artificial Intelligence
Previous article Next article
feedback
Top