We are Team Data at the Office for Statistics and Data of the Canton of Zurich. We provide data science expertise and use artificial intelligence (AI) to support cantonal staff and improve public services. AI is one of several tools the canton uses in its digital transformation.
We work with other cantonal teams to build prototypes and proofs of concept. Our pilots test whether machine learning can solve specific problems in administrative workflows and help staff build expertise in using AI.
We share what we learn, along with code and data, on GitHub whenever possible. For feedback, email us or open an issue or pull request in the relevant repository.
How the Canton of Zurich uses AI
- TranscriboZH. Transcribe audio and video files, then view and edit transcripts in a standalone HTML editor.
- Hybrid Search for the State Archives. Search large document collections by keyword and meaning. Search the Zurich State Archives' central collections.
- Document Research Tool. Find documents by keyword and meaning, then choose which ones to discuss with a large language model (LLM).
- Deep Research. Run automated, multi-step research and analysis on your own document collection and generate a report.
- AI Chat. Run an LLM chat application with document processing on your own infrastructure.
- Semantic Search Evaluation Tool. Evaluate semantic search with custom datasets, metrics, and embedding backends.
- Hybrid Search Evaluation Tool. Benchmark embedding models with BM25 keyword search and vector search on your own data or other MTEB-compatible datasets. The tool reports MRR@K and Hit Rate@K, plus embedding latency and memory use where available.
- Named Entity Recognition (archived). Use this framework to identify names and other entities in administrative text.
- ZIX Understandability Index. Estimate how easy a German text is to understand.
- Language Simplification Tool. Rewrite German administrative texts in Einfache Sprache or Leichte Sprache.
- Language Simplification API. Integrate German text simplification into other services.
- Glossarizer. Use LLMs to create plain-language glossaries from texts, URLs, or files.
- Alt Text Generator. Automatically generate German alt text for web images using AI.
- Subtitle to Speech Tool. Turn subtitle files into German audio that follows the subtitle timings.
- OGD AI Analyzer. Check the quality of metadata in catalogs that use the Data Catalog Vocabulary (DCAT).
- OGD MetaFairy. Draft dataset descriptions with AI.
- OGD AI Search. Search an open data catalog's metadata by keyword and meaning across languages.
- Plausi R package. Estimate expected voting results and flag statistical anomalies. The package powers PlausiApp, which several cantons use to check reported results. Request access to PlausiApp.
- machinelearningZH. AI and machine learning in the Canton of Zurich (this profile).
- statistikZH. Office for Statistics and Data of the Canton of Zurich.
- openZH. Specialist Unit for Open Government Data of the Canton of Zurich.