Abstract
The biggest struggle that teams have in the current artificial intelligence (AI) arms race is attaining high-quality, representative data to train or test product solutions. Prompts to a Large Language Model (LLM) must be increasingly complex, and even then, models struggle to generate conversational data that could reasonably pass for actual conversation between two or more people. By using an iterative prompting method that builds in real-world variation, a system is proposed herein that provides for the ability to create more realistic conversational data.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Tomlinson, Andrew; Kraus, Kelsey; and Maiya, Rupa, "SYNTHETIC GENERATION OF REALISTIC CONVERSATIONAL DATA USING LARGE LANGUAGE MODELS", Technical Disclosure Commons, (June 21, 2024)
https://www.tdcommons.org/dpubs_series/7118