Contrastive Speaker-Aware Learning for Multi-party Dialogue Generation with LLMs

2025-03-11

AI Summary

Existing approaches to multi-participant dialogue modeling face two key limitations: (1) the complexity of speaker interactions and (2) reliance on manually annotated dialogue relations, which constrains scalability and generalizability. To address these, we propose Speaker-Attentive LLM (SA-LLM), a pretrained large language model enhanced with speaker-aware encoding and unsupervised contrastive learning. SA-LLM implicitly captures speaker consistency and contextual coherence—without requiring explicit dialogue relation annotations—via three core components: speaker-identity embeddings, contrastive loss optimization, and context-aware decoding. Evaluated on Ubuntu IRC and Movie Dialogues, SA-LLM achieves new state-of-the-art performance: automatic metrics (e.g., BLEU, ROUGE, BERTScore) show significant gains, while human evaluation confirms consistent improvements in fluency, coherence, informativeness, and response diversity. This work represents a principled advance in modeling dynamic, multi-role interactions in open-domain dialogues.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Conversational AI/Dialog Systems
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Multi-party dialogue generation presents significant challenges due to the complex interplay of multiple speakers and interwoven conversational threads. Traditional approaches often fall short in capturing these complexities, particularly when relying on manually annotated dialogue relations. This paper introduces Speaker-Attentive LLM (SA-LLM), a novel generative model that leverages pre-trained Large Language Models (LLMs) and a speaker-aware contrastive learning strategy to address these challenges. SA-LLM incorporates a speaker-attributed input encoding and a contrastive learning objective to implicitly learn contextual coherence and speaker roles without explicit relation annotations. Extensive experiments on the Ubuntu IRC and Movie Dialogues datasets demonstrate that SA-LLM significantly outperforms state-of-the-art baselines in automatic and human evaluations, achieving superior performance in fluency, coherence, informativeness, and response diversity. Ablation studies and detailed error analyses further validate the effectiveness of the proposed speaker-attentive training approach, highlighting its robustness across different speaker roles and context lengths. The results underscore the potential of SA-LLM as a powerful and annotation-free solution for high-quality multi-party dialogue generation.

Problem

• Addresses multi-party dialogue generation challenges
• Leverages speaker-aware contrastive learning strategy
• Improves fluency, coherence, and response diversity

Innovation

• Speaker-Attentive LLM with contrastive learning
• Pre-trained LLMs for multi-party dialogue generation
• Annotation-free speaker role and context learning