Host: The Japanese Society for Artificial Intelligence
Name : The 100th SIG-SLUD
Number : 100
Location : [in Japanese]
Date : February 29, 2024 - March 01, 2024
Pages 39-46
In this study, we focused on the relationship between the experience of radio personalities and the characteristics of disfluency elements appearing in anecdotal talks, in order to elucidate their features. Specifically, we classified disfluency elements into five categories: "fillers," "self-repairs," "phonemic elongations," "voiceless pauses," and "breathing (exhalation/inhalation)," and analyzed their occurrence frequencies in talks by both experienced and novice personalities. As a result, the overall frequency of disfluency elements was higher in talks by novices than in those by experienced personalities. "Elongations" and "fillers" were more dominant in novice personalities, while the frequency of "self-repairs" was lower in talks than by experienced personalities. Furthermore, regarding the occurrence patterns of fillers, it was evident that in novice talks, specific fillers tended to be biasedly used and fillers sometimes occurred consecutively. Therefore, it can be inferred that minimizing the use of disfluency elements and avoiding a bias towards specific types of fillers are strategies for effective communication by radio personalities.