All Topics
All Topics
Technology
Technology
Design
Design
Programming
Programming
Science
Science
News
News
Gaming
Gaming
Entertainment
Entertainment
Business
Business
Finance
Finance
Sports
Sports
Health
Health
Food
Food
Travel
Travel
Art
Art
Music
Music
Books
Books
Education
Education
Politics
Politics
Personal
Personal
No algorithm. No AI slop. No ads. Just RSS. Pro-human. Indie writers. Real journalism. Open web. Chronological. Hand toasted.

GuidaPA: A Federated Learning-Based Privacy-Preserving Chatbot for Italian Public Administration

By

[Submitted on 31 May 2026]

8d ago· 2 min readenInsight

Summary

GuidaPA is a privacy-preserving chatbot for the Italian Public Administration trained via Federated Learning (FL) on documentation from two national PA platforms (SIGESON and SIDFORS). The system uses QLoRA (4-bit) over 15 federated rounds with an 80/20 train-test split, integrating role-based access control, secure client-side preprocessing, and monitoring of non-IID effects. The best federated model achieves ROUGE-1/2/L of 61.10/55.77/59.44, BLEU-4 of 45.02, and METEOR of 63.94, approaching private centralized fine-tuning performance while keeping data on-site. Domain fine-tuning significantly improves over the general-purpose baseline, with ROUGE-1 rising from 41.45 to 62.18 and BLEU-4 from 26.97 to 50.90.

Key quotes

· 4 pulled
The intended deployment extends to restricted internal sources (e.g., tickets, officer manuals, database extracts) that can not be centrally pooled due to regulatory and organizational constraints.
The best federated model achieves ROUGE-1/2/L of 61.10/55.77/59.44, BLEU-4 of 45.02, and METEOR of 63.94-close to private centralized fine-tuning while keeping data on-site.
Overall, the results indicate that FL can deliver high-quality conversational AI for public services without centralized data sharing
Domain fine-tuning improves ROUGE-1 from 41.45 to 62.18 and BLEU-4 from 26.97 to 50.90.
Snippet from the RSS feed
We present GuidaPA, a privacy-preserving chatbot for the Italian Public Administration (PA) trained via Federated Learning (FL) on documentation from two national PA platforms, SIGESON and SIDFORS. Our corpus includes approximately 8 pages of SIGESON manu

You might also wanna read