Unraveling Emotional Dynamics in Conversations with Swarm Decomposition, Affect Dynamics, and Machine Learning

Conversational Emotional Climate Recognition (CERC) endeavors to uncover the subtle emotional undercurrents inherent in spoken interactions, offering profound insights into the intricate fabric of social dynamics. This paper introduces MLSwarm, an innovative CERC method that seamlessly integrates machine learning and swarm decomposition (SWD). MLSwarm meticulously applies swarm decomposition to audio signals, extracting components, and subsequently derives Mel-frequency cepstral coefficients (MFCC) from these components, yielding nine distinctive features. Evaluated in two configurations, MLSwarm attains noteworthy accuracy: 76.7A/65.3V and 79.2A/73.6V. While SWD empowers ML-Swarm to unlock intricate patterns within audio signals, it's the integration of affect dynamics (AD) that truly elevates its performance. This potent combination not only amplifies MLSwarm's accuracy in deciphering emotional tones but also grants deep insights into the delicate emotional dynamics underlying social interactions. The findings emphasize MLSwarm's effectiveness in capturing the emotional tone of conversations between peers, enriching our comprehension of the emotional dynamics at play in social interactions.