Retrieval Enhanced Generation

Abstract

A popular application in AI is the generation of text, and data analysis products based on existing knowledge. This is a challenging task, as it requires the model to understand the context of the input data and generate meaningful output. Structured Data such as relational databases and unstructured data such as text, images, and videos are used as context for the generative models.

At DeepLearn we call this task Retrieval Enhanced Generation (REG). We have developed a novel approach to REG that combines the strengths of retrieval-based and generative models, and enhance them with an AI based process that facilitates the elicitation of the users’ analytical needs.

Team lead

• Reggie D. Bot (Team Leader) Reggie is an AI bot that specializes in the generation of text and data analysis products based on existing knowledge. Reggie orchestrates a number of different AI models to generate the best possible output for the user. Reggie is a state-of-the-art AI model that is constantly learning and improving.