A Demo of GeoQA^3: Towards An Accessible AI-based Question-Answering System for Geoanalytics

  • Chu Li ,
  • ,
  • Arnavi Chheda-Kothary ,
  • Ather Sharif ,
  • Henok Assalif ,
  • Jeffrey Heer ,
  • Jon E. Froehlich

Proceedings of the 27th International ACM SIGACCESS Conference on Computers and Accessibility |

Publication

Geovisualizations are powerful tools for analyzing and interpreting spatial data; however, they are historically inaccessible to screen-reader users. We introduce GeoQA3, an Accessible AI-based Question-Answering (QA) system to enable blind users to perform geo-spatial analytics. GeoQA3 relies on a custom QA pipeline that combines map interactions with chat questions to form queries and uniquely combines geo-statistical analysis with LLM-based summaries. In a remote lab study with six screen-reader users, we found that participants successfully employed diverse querying strategies for spatial analysis and highly valued the AI Chat component for its interactive responses. During the ASSETS demo session, attendees will use GeoQA3 to explore two key questions, mirroring our user study: potential biases in digital access in the US and how energy sources differ geographically across the US.