Examining Synthetic Speech in Corporate Training: Focusing on Impression Formation and Likability

Abstract

This study aims to identify the characteristics of synthetic speech that tend to elicit higher levels of likability by examining evaluations and impressions of synthetic speech. To achieve this objective, two practices were conducted. Practice 1 involved a confirmation test, subjective evaluation, and an impression formation survey on video teaching materials featuring an instructor's human voice and a synthetic speech. Practice 2 focused only on evaluating synthetic speech (four types in total: two male and two female), to survey on evaluating and impression formation surveys. In Practice 1, the results indicated no significant differences in test scores or subjective evaluations between the two materials.  However, in the impression formation survey, human voices were significantly more highly evaluated for characteristics such as “Friendly,” “Pleasant,” “Sensible,” and “Approachable.”  In Practice 2, synthetic speech perceived as “Pleasant,” “Approachable,” and “Kind” tended to receive higher evaluation scores.