Understanding Emotion Loss and Correction Mechanisms in Speech-to-Text Communication

Speech-to-text(STT) systems have limitations in conveying emotions in remote communication because they cannot sufficiently reflect nonverbal nuances such as intonation, emphasis, and tone. This study aims to explore user perceptions of emotional loss during STT conversion and strategies for correcting it. We conducted an STT-based message writing experiment using six emotional scenarios and semi-structured interviews with 15 participants who had experience using STT. The analysis revealed that most participants perceived emotions as neutralized in the STT conversion results and employed correction strategies such as symbols, emojis, and ironic expressions, considering contextual factors such as emotional intensity, social relationships, and message interpretability. This study reveals that users' emotional correction strategies go beyond simple error correction to reflect social context in proactive interaction, which could contribute to establishing practical design criteria for future emotion-friendly STT system design.