EmotiCon: Enhancing Public Speaking Training with Context-Aware Feedback on Emotional Delivery

Abstract

Effective public speaking relies on aligning emotional delivery with semantic content. While existing digital coaches analyze paralinguistic cues like tone and pitch, they often neglect the context of what is being said. This paper introduces EmotiCon, a novel system that offers context-aware emotional feedback by analyzing a speaker’s audio, transcribed text, and preparation materials. EmotiCon segments speech, detects emotion across modalities, and compares it with relevant contextual passages to assess emotional congruence. It then generates targeted recommendations using a Large Language Model. A user study showed that context-aware feedback was better received than simple feedback. Quantitative evaluation using the extended Technology Acceptance Model (TAM) followed by semi-structured interviews revealed strong scores. Our findings suggest that integrating semantic context into feedback generation makes digital speech coaching tools more actionable and effective for learners.