The Emotional Portrayal of an Ordinary Talk

Authors: Neda Mousavi, Felix Burkhardt

Abstract:

This study examines how automatic emotion recognition models capture emotional expression in ordinary speech. Using German autobiographical narratives, the results of linguistic and acoustic models are compared at both the speaker and segment levels. While linguistic emotionality is relatively homogeneous and expressed locally, acoustic emotionality shows greater variability between speakers and stronger temporal persistence. The results suggest that dynamic patterns in model-derived emotionality are modality-dependent and reflect the different roles of lexical content and prosody in ordinary speech.