DiaSynth: Synthetic Dialogue Generation Framework for Low Resource Dialogue Applications
This addresses data limitations for dialogue system developers, offering a scalable alternative to traditional data collection, though it is incremental as it builds on existing LLM and CoT methods.
The paper tackles the scarcity of domain-specific dialogue datasets by introducing DiaSynth, a synthetic dialogue generation framework using LLMs and CoT reasoning, which improves dialogue summarization by 16.47% over base models and captures 90.48% of in-domain data performance.
The scarcity of domain-specific dialogue datasets limits the development of dialogue systems across applications. Existing research is constrained by general or niche datasets that lack sufficient scale for training dialogue systems. To address this gap, we introduce DiaSynth - a synthetic dialogue generation framework capable of generating high-quality, contextually rich dialogues across a wide range of domains. Unlike existing frameworks, DiaSynth uses Large Language Models (LLMs) and Chain of Thought (CoT) reasoning to generate dynamic, domain-specific dialogues with simulated personas and diverse conversational features. We perform our experiments by generating synthetic data using different LLMs and few-shot examples from DialogSum and SAMSum. The pretrained language models fine-tuned on the synthetic data outperform the base models by 16.47% on dialogue summarization, while the comparison between models fine-tuned on in-domain data and synthetic data shows that the synthetic data is able to capture 90.48% of the performance distribution of the in-domain data on dialogue summarization. The quality of the data generated also increases as we increase the size of LLM from 3B to 8B. These results validate DiaSynth's potential as a robust alternative to traditional data collection methods. We open source the code and data generated for future research.