Dataset generation using large language models

This invention describes a system to automatically create large sets of natural language questions for training AI chatbots. It works by taking a basic question template, filling in blanks with specific words, and then asking a large language model (like ChatGPT) to generate a full, natural-sounding question. These generated questions are then used to teach another AI model how to have conversations in a specific area. The claims narrow this to templates that include an example question, its expected answer, and placeholders for modification.