Investigating Cross-Cultural Vocal Emotion Recognition With an Affectively and Linguistically Balanced Design.

Abstract

This study investigates cross-cultural vocal emotion recognition in a corpus with an affectively and linguistically balanced design. It has two main goals, one theoretical and the other methodological. First, it aims to explore the recognition of emotions in two typologically different languages, Dutch and Korean, within and across cultures. Second, it aims to contribute to the methodological development of the study of cross-cultural vocal emotion recognition by presenting a new corpus for Dutch and Korean emotional speech (the Demo/Koremo corpus), containing portrayals of eight emotions differing in arousal, valence, and basicness (joy, pride, tenderness, relief, anger, fear, sadness, irritation) produced by Dutch and Korean actors, and communicated in a single pseudo phrase which was viable in both languages. Dutch and Korean participants listened to recordings of all emotions produced by the Dutch and Korean actors and indicated for each one which emotion they thought it expressed. Both groups of listeners recognized emotions significantly above chance in both languages, but more accurately in their native language, in line with the Dialect Theory of emotion. Low-arousal emotions, negative emotions, and basic emotions were recognized more accurately than their counterparts. While some of these results replicate earlier findings, others-the effect of arousal and the within-cultural effect of valence and basicness-had not been previously investigated. This study provides new insights in cross-cultural vocal emotion recognition and contributes to the methodological toolkit of intercultural emotion recognition research.