Handwriting and Text Recognition

Services
Author

Digital Humanities

The Digital Humanities at the University of Bern offers different solutions for text recognition

TrOCR

For large scale data processing (e.g. within the FLOW project), we use Transformer OCR. For this we provide support on demand and publish our models open source: huggingface.co/dh-unibe/models.

eScriptorium

eScriptorium is a web application offering a workspace to manage the various steps of a transcription campaign.

eScriptorium is open source but needs to be deployed on a server infrastructure. For the FLOW-project, the University of Bielefeld offers an eScriptorium instance.

Please get in touch if you would like to use eScriptorium.

Transkribus

Transkribus is an machine learning-based platform for the automated recognition and transcription of handwritten and historical documents. It is widely used in the humanities and archives to digitize and make accessible historical texts.

The University of Bern (through the Walter Benjamin Kolleg) is member of the READ Cooperative, which is responsible for the maintenance of Transkribus.

Transkribus enables users to: - Automatically transcribe handwritten documents using pre-trained AI models or custom-trained models tailored to specific handwriting styles. - Manually transcribe and correct texts within its user-friendly interface. - Train custom AI models to recognize specific handwriting, which is particularly useful for unique or challenging scripts. - Search and annotate documents, facilitating detailed analysis and research. - Publish collections online through Transkribus Sites, making them accessible to a broader audience.

Contact

For support and further information, please contact us at digitalhumanities@unibe.ch.

Back to top