The OCR-D coordination project as well as in four implementation and three module projects will continue to improve the OCR-D software in the coming years, paving the way for the mass digitisation of all works in the VD 16, 17 and 18.
Since 2015, the DFG funding initiative has been working on the further development of Optical Character Recognition (OCR) processes for historical prints and has already been able to present a prototypical OCR software [1]. With its multitude of available tools, this software is suitable for a wide range of different application scenarios. In phase III, the focus is on the conceptual preparation for the automatic generation of full texts for VD 16, VD 17 and VD 18.In addition, four implementation projects are working on integrating OCR-D into existing applications and infrastructures, while three module projects are further optimising OCR-D tools.
On 30 July, our kick-off workshop took place, ushering in Phase III of OCR-D. The coordination project team gave an introduction to OCR-D's goals and public communication channels [2], the OCR-D software status and plans [3] and web API [4], and how to handle Ground Truth data in OCR-D [5]. In addition, the coordination project gave insights into the previous practice of software development in OCR-D [6] including opportunities to contribute. Furthermore, the implementation and module projects presented themselves to the interested community and our cooperation partners in short presentations [7].
UB Braunschweig, SLUB Dresden and UB Mannheim are extending OCR-D and Kitodo for productive mass digitisation; SUB Göttingen and GWDG are working on performance optimisation by deploying OCR-D on a High-Performance Cluster; GEI Braunschweig, HCI and ZPD of the University of Würzburg will make OCR-D tools available in OCR4all; ULB Sachsen-Anhalt will implement OCR-D in its Open Source mass digitisation infrastructure. While these project partners are working on four implementation scenarios, three projects will improve the OCR-D modules from Phase II: UB Mannheim is developing work-specific training with Tesseract and Calamari; JGU Mainz and FAU Erlangen-Nuremberg are improving font group recognition for better-fitting OCR models; and the SUB Göttingen and GWDG project is optimising the reliability, searchability and granular referencing of OCR results in the long-term archive OLA-HD.
In our chat channel, the Gitter lobby [8], we always keep you up to date on public OCR-D events and you are welcome to join our community. For more information on how you can get in touch and contribute to OCR-D, please visit our website [9]. We will also let you know in the Gitter lobby and on our website once you can subscribe to our newsletter, which will be launched soon.
[1]https://github.com/OCR-D
[2]https://ocr-d.de/assets/kick-off/phase3.pdf
[3]https://ocr-d.de/assets/kick-off/spec_core_ocrd_all.pdf
[4]https://ocr-d.de/assets/kick-off/web-api.pdf
[5]https://ocr-d.de/assets/kick-off/gt.pdf
[6]https://ocr-d.de/assets/kick-off/software-development.pdf
[7]https://ocr-d.de/assets/kick-off/lightning-talks.pdf
[8]https://gitter.im/OCR-D/Lobby
[9]https://ocr-d.de/en/platforms