DIGILOG at ECAI 2025

We are pleased to announce that the paper "Query Carefully: Detecting the Unanswerables in Text-to-SQL Tasks", authored by ZHAW researchers Jasmin Saxer, Andreas Weiler, and Kurt Stockinger, together with UZH researchers Isabella Maria Aigner and Luise Linzmeier, has been accepted at the International Joint Workshop on Artificial Intelligence for Healthcare (HC@AIxIA) and HYbrid Models for Coupling Deductive and Inductive ReAsoning (HYDRA) 2025. The workshop was co-located with the 28th European Conference on Artificial Intelligence (ECAI 2025).

The work was presented by Jasmin Saxer at the workshop in Bologna, Italy.

While the paper is evaluated in a medical Text-to-SQL setting, its core contribution focuses on detecting when a natural language query cannot be answered based on the available data. This challenge is highly relevant for Digilog. In local government contexts, users often interact with complex and incomplete public datasets, where ambiguous or out-of-scope questions can easily lead to misleading results.

By enabling explicit no-answer detection and transparent query handling, this research supports Digilog's goal of developing trustworthy, logic-based access to public sector data and helping users better understand both the capabilities and limitations of AI-driven data access.