Learn Less, Parse More: Using Grammars To Post-Process Generated Text, Marie Chatfield Rivas

Home
My videos
Learn Less, Parse More: Using Grammars To Post-Process Generated Text, Marie Chatfield Rivas

Summary

Marie Chatfield Rivas, a software engineer at data.world, discusses the importance of parsing text generated by large language models (LLMs) to improve reliability and correctness. They explain how to use standard grammars and post-processing techniques to transform LLM responses. Rivas provides examples of using parsers to generate SQL queries, custom search syntax, and RDF models with a strict ontology. They also discuss optimizing grammar for LLM cooperation and supporting multiple languages with post-processing. Rivas demonstrates these concepts using a recipe project and explains how structured data can be imported into a data catalog platform.

Have questions about the video? Create a FREE account to ask Wiz AI

Sign up

7 days free trial