eScriptorium is an open-source platform for the segmentation and transcription of historical manuscripts, printed books, and archival documents. It provides a web-based interface for Handwritten Text Recognition (HTR) and Optical Character Recognition (OCR), enabling users to generate automatic transcriptions and perform manual corrections within a collaborative environment.
The platform is designed to manage the full transcription workflow, from image ingestion and layout analysis to the training of dedicated recognition models and the export of structured data.
Design Philosophy
The architecture of eScriptorium is built around the specific requirements of the humanities and the complexity of historical documents.
Script and Layout Agnosticism
eScriptorium is not optimized for a single language or period. Instead, it is language-agnostic and largely built upon Kraken, a turn-key OCR/HTR engine. This architecture allows the platform to process diverse scripts, including Latin, Hebrew, Arabic, Syriac, Greek, and Coptic, as well as complex layouts containing marginalia, interlinear glosses, and irregular columns common in manuscript traditions.
The Human-in-the-Loop Workflow
The platform operates on the principle that high-accuracy transcription of historical sources requires iterative training, integrating the human-in-the-loop methodology directly into the user interface. The workflow typically begins with the user manually segmenting and transcribing a subset of pages. These samples are then used to train a specific model directly on the server, which is subsequently applied to predict text on the remainder of the document. Finally, the user validates or corrects the output, creating new ground truth data to further refine the model.
Interoperability and Open Standards
To ensure sustainability and integration within the wider digital ecosystem, eScriptorium relies on established community standards:
- IIIF Compliance: Images can be imported via IIIF manifests.
- Standard Exports: Data is exportable in ALTO XML and PAGE XML formats.
- Open Source: The software is published under the MIT License.
Project History
eScriptorium was initiated in 2018 within the framework of the Scripta PSL research program (“History and Practices of Writing”) at the Université PSL (Paris Sciences et Lettres). The project originated from the need for a graphical user interface that would make the command-line functionalities of the Kraken engine accessible to humanities scholars.
The initial development was driven by the specific needs of philologists working with non-Latin scripts and complex layouts, areas where commercial OCR solutions often remain insufficient. Since its inception, the project has expanded through funding from various sources such as the RESILIENCE research infrastructure and the Biblissima+ project.
Governance and Maintainers
The development and maintenance of eScriptorium are currently primarily led by a partnership between two French research institutions, the Digital Humanities group at the École Pratique des Hautes Études (EPHE) and the ALMAnaCH project-team (Automatic Language Modelling and ANalysis & Computational Humanities) at Inria.
Steering Committee
Strategic decisions regarding the platform’s long-term roadmap, scientific direction, and sustainability are made by a steering committee composed of representatives from the founding and maintaining organizations.
Current Members:
- Daniel Stoekl Ben Ezra (EPHE-PSL, UMR AOROC 8546)
- Peter Stokes (EPHE-PSL, UMR AOROC 8546)
- Benjamin Kiessling (Inria)
- Mathew Barber (Aga Khan University, Institute for the Study of Muslim Civilisations)
- David Smith (Northeastern University)
- Thibault Clérice (Inria)
- Hassen Aguili (Inria)