Bullinger Digital
The project aimed to digitally process the correspondence of the Zurich reformer Heinrich Bullinger (1504–1575) and was based on the historical-critical edition of the Bullinger correspondence (HBBW) and the information compiled by the HBBW project. The processed sources of knowledge were made accessible in an online search system, together with specially created digital reproductions of the manuscripts. To facilitate searching, the texts written in Early New High German were normalized. In addition, a system for machine translation from Latin to German was developed. Methods for automatic handwriting recognition were also trained and compared in order to accurately access the letters that had not yet been transcribed. "Bullinger digital" was led by the Institute of Computational Linguistics together with the Institute for Swiss Reformation History. The University of Bern, the University of Applied Sciences Fribourg, as well as the Zurich State Archives and the Zurich Central Library, were project partners.
Technical Tools
- Database / Search System (Bullinger Digital): SQLite, Python, Cantaloupe IIIF Server, Elasticsearch, React (NextJS)
- Database / Search System (Phase 2): TEI Publisher
- Automatic Handwriting Recognition: HTR-Flor, Transkribus, TrOCR
- Machine Translation (Bullinger Digital): Sockeye, Classical Language Toolkit
- Machine Translation (Phase 2): GPT-4o
- Custom-built tools for language identification (Latin vs. Early New High German) and code-switching detection
- Custom-built app for assigning scans
- CMS for manual correction of metadata: Directus
- "Contribute" tool (custom-built) for the citizen science campaign for named entity recognition and linking: NextJS and PostgreSQL; source code on GitHub
- XML Editor Oxygen
- GitHub for version control and downloads
- Trello for project management
Usage Options
The TEI XML data is available via the GitHub repository of the Zurich State Archives at this link (CC BY NC ND 3.0).
Documentation on models and tools is available via this GitHub link.
English translations of all abstracts (Regesten) and manually transcribed letters, created using LLMs, can be accessed via this GitHub link.
The source code for the "Contribute Tool" is available via this GitHub link.
Project Responsibility
Prof. Dr. Martin Volk, Institute of Computational Linguistics
Prof. Dr. theol. Tobias Jammerthal, Institute for Swiss Reformation History
Prof. em. Dr. theol. Peter Opitz, Institute for Swiss Reformation History