Leveraging Graph Structures and Large Language Models for End-to-End Synthetic Task-Oriented Dialogues

TL;DR

GraphTOD is a user-friendly framework that uses graph structures and large language models to generate high-quality task-oriented dialogues, reducing the cost and complexity of dataset creation.

Contribution

We propose GraphTOD, an end-to-end system that simplifies dialogue generation through transition graphs, making it accessible without custom prompts or coding.

Findings

High-quality dialogues generated across multiple domains

Significant reduction in dataset creation costs

Accessible to non-technical users

Abstract

Training task-oriented dialogue systems is both costly and time-consuming, due to the need for high-quality datasets encompassing diverse intents. Traditional methods depend on extensive human annotation, while recent advancements leverage large language models (LLMs) to generate synthetic data. However, these approaches often require custom prompts or code, limiting accessibility for non-technical users. We introduce GraphTOD, an end-to-end framework that simplifies the generation of task-oriented dialogues. Users can create dialogues by specifying transition graphs in JSON format. Our evaluation demonstrates that GraphTOD generates high-quality dialogues across various domains, significantly lowering the cost and complexity of dataset creation.