Dialogue Summarization via Topic and Participant Decomposition

This framework uses an adapted hierarchical Chain-of-Agents approach to summarize dialogues by modeling semantic content and emotion dynamics. It decomposes inputs into topic segments and participant-specific utterances to track emotional trajectories across interactions.