AI in Swedish Can Transform the Public Sector

RISE is developing an AI language model trained on Swedish texts from the National Library of Sweden’s collection dating back to 1661. This initiative aims to help government agencies manage growing volumes of emails, reports, and documents more efficiently.

The Project

The “Language Models for the Swedish Government” project creates a digital comprehension model adapted specifically for government communications. The project brings together key partners:

• Swedish Public Employment Service
• Swedish Tax Agency
• Swedish Agency for Economic and Regional Growth
• AI Sweden
• Lulea University of Technology
• Peltarion

Capabilities

According to Magnus Sahlgren, Head of Natural Language Processing at RISE: “AI-based text categorisation systems can be developed much more quickly than before” using this Swedish language foundation.

The system enables:

• Automatic email routing to the right department
• Document categorization based on content understanding
• Context-aware responses to citizen inquiries without requiring exact terminology matching

Quality and Ethics

The model was trained using editorial text rather than social media content, ensuring a higher quality baseline. Rigorous testing addresses potential biases and regional language variations across Sweden before deployment in public services.

Expected Benefits

The automation of document management promises significant improvements:

• Reduced stress on government employees
• Freed time for skilled, human-centric work
• Improved and faster citizen services
• More consistent handling of government communications

A Swedish Approach

Unlike relying on foreign language models, this initiative ensures that Swedish authorities have access to AI tools specifically adapted for the Swedish language and administrative context, supporting digital sovereignty in critical public sector functions.