Reading Between the Lines: The One-Sided Conversation Problem

2025-11-04

AI Summary

This paper addresses the practical constraint in privacy-sensitive domains—such as telemedicine and call centers—where only unidirectional dialogue transcripts (i.e., one speaker’s utterances) are available, formalizing it as the One-Sided Conversation (1SC) problem. It focuses on two core tasks: missing-turn reconstruction and one-sided dialogue summarization. Methodologically, it introduces a novel one-sided dialogue paradigm, demonstrating that incorporating the next-turn content and length information significantly improves reconstruction fidelity; proposes a placeholder-based prompting mechanism to mitigate LLM hallucination; and—crucially—establishes for the first time that high-quality summaries can be generated *without explicit turn reconstruction*. The approach integrates prompt engineering with model fine-tuning. Experiments on MultiWOZ, DailyDialog, and Candor show that large models achieve reasonable reconstruction via prompting alone, while smaller models require fine-tuning; remarkably, the summarization task achieves state-of-the-art performance even under zero-reconstruction conditions.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Conversational AI is constrained in many real-world settings where only one side of a dialogue can be recorded, such as telemedicine, call centers, and smart glasses. We formalize this as the one-sided conversation problem (1SC): inferring and learning from one side of a conversation. We study two tasks: (1) reconstructing the missing speaker's turns for real-time use cases, and (2) generating summaries from one-sided transcripts. Evaluating prompting and finetuned models on MultiWOZ, DailyDialog, and Candor with both human A/B testing and LLM-as-a-judge metrics, we find that access to one future turn and information about utterance length improves reconstruction, placeholder prompting helps to mitigate hallucination, and while large models generate promising reconstructions with prompting, smaller models require finetuning. Further, high-quality summaries can be generated without reconstructing missing turns. We present 1SC as a novel challenge and report promising results that mark a step toward privacy-aware conversational AI.

Problem

• Infer missing dialogue turns from one-sided conversations
• Generate summaries directly from partial conversation transcripts
• Address privacy constraints in real-world conversational AI applications

Innovation

• Inferring missing dialogue turns from one-sided conversations
• Using placeholder prompting to reduce hallucination in reconstructions
• Generating summaries directly from partial transcripts without reconstruction