LLMs are notoriously bad at answering quantitative questions. The Italian open data community is using them as natural-language based interfaces to authoritativ
The Italian open data community is leveraging large language models (LLMs) to create natural language interfaces for authoritative data sources, despite LLMs' known limitations with quantitative questions. This advancement is facilitated by established standards for data and metadata publication, such as SDMX for statistical data and DCAT for open metadata. Key contributors to this effort include Alberto Borruso and Matteo Fortini. AI
IMPACT Enables more accessible querying of structured data, potentially increasing data utilization.