Public AI Assistant to Worldwide Knowledge: Answering Queries of Structured and Unstructured Data

Explore a 34-minute talk from Stanford University that introduces Genie, a public AI assistant developed by the Stanford Open Virtual Assistant Lab. Learn how Genie processes natural language queries to retrieve information from hybrid data sources containing both structured and unstructured content. The presentation explains how Genie translates user requests into formal queries using SQL, SPARQL, and the novel SUQL language, which extends SQL to integrate information retrieval from free text with relational database concepts. See real-world applications through case studies of Genie's deployment at Wikidata and the Knight Election Hub (using Federal Election Commission data). Speaker Shicheng Liu demonstrates how this agentic approach combines evaluation of intermediate queries and knowledge graph navigation to generate comprehensive responses, representing an important advancement for public access to worldwide knowledge. This workshop was recorded on February 14, 2025, at Stanford University.