Par Daniel Stoekl Ben Ezra (Directeur d'études, EPHE-PSL, UMR 8546 AOrOc). Rendez-vous IIIF360, un événément en ligne autour des standards et technologies IIIF organisé par le consortium IIIF360 (Biblissima, Campus Condorcet, Huma-Num) le 24 mars 2021 : https://projet.biblissima.fr/fr/evenements/rendez-vous-iiif360-2021

1. eScriptorium: An Open Source
Platform for
Historical Document Analysis
Daniel Stökl Ben Ezra
Peter Stokes
Marc Bui
Ben Kiessling
Robin Tissot

2. eScriptorium
• Blog: http://escripta.hypotheses.org
• UI Code: https://gitlab.inria.fr/scripta/escriptorium
• AI Code: https://github.com/mittagessen/kraken
Funded by: PSL IRIS Scripta, H2020 Resilience, MENESR, DIM STCN Ile de
France, EquipEx Biblissima+, [indirectement: Mellon, MCC]

3. eScriptorium Universe
Scripta PSL:
eScriptorium
LectauRep
INRIA
ANF
openITI
North-Eastern
U Maryland
ERC
Vietnamica
EPHE
U-Bib
Heidelberg
?
National
Library of
Israel ?
ENC
Sorbonne
Université
DIM STCN
Observatoir
e de Paris
IRHT
H2020
Resilience
• manuscriptologIA
High Performance
Computing Cluster at
mesoPSL
Biblissima+
TGIR Huma-Num

7. current
• Import:
IIIF
pdf, imgfiles (jpg, png, …), alto, PageXML,
trained segmentation or transcription models
• Ergonomic UI for manual segmentation, transcription and (soon) annotation.
4 panels (facsimile, segmentation, transcription, text-annotation)
( user definable architectures)

8. ↓ Metadata imported via iiif

9. current
• Import: IIIF, pdf, imgfiles (jpg, png, …), alto, PageXML,
trained segmentation or transcription models (user definable architectures)
• Ergonomic UI for manual segmentation, transcription and (soon) annotation.
4 panels (facsimile, segmentation, transcription, text-annotation)

10. Ergonomic transcription e.g. of vertical or oblique lines

11. BL ms Add. 27296
Transcription font size automatically adapted to manuscript line

12. current
• Import: IIIF, pdf, imgfiles (jpg, png, …), alto, PageXML,
trained segmentation or transcription models (user definable architectures)
• Ergonomic UI for manual segmentation, transcription and (soon) annotation.
4 panels (facsimile, segmentation, transcription, text-annotation)
• Automatic segmentation (lines, semantic lines and regions, also overlapping)
based on user-defined ontologies.
• Automatic transcription according to the principles set by the user.
• Export: alto 4(!), PageXML, txt, imgfiles (jpg, png ,…)
trained segmentation or transcription models
• Powerful and growing API

13. Segmentation and Transcription
Demonstration
↑ User definable
segmentation ontology

15. Locate illuminations through layout segmentation

16. Automatic segmentation result of ms specific model

17. Ergonomic correction

19. Jbaiter Mirador textoverlay plugin

20. eScriptorium (near) FUTURE
Scripta PSL:
eScriptorium
LectauRep
INRIA
ANF openITI
North-Eastern
U Maryland
ERC
Vietnamica
EPHE
U-Bib
Heidelberg
?
National
Library of
Israel ?
ENC
Sorbonne
Université
DIM STCN
Observatoir
e de Paris
IRHT
H2020
Resilience
• Search
• Trainable reading order
• Prototype for text annotation (NE,
ecdotic) with TEI-Export
• Prototype for image annotation
(e.g. Digipal / Archetype)
• manuscriptologIA
High Performance
Computing Cluster at
mesoPSL
• Customizable virtual
keyboard
• Vertical interface for Chinese
• Automatic textalignment
• Additional simplified interface
• Improved project management
• Crowdsourcing interface
Biblissima+
TGIR Huma-Num

21. Transcription created automatically without
specific transcription BnF syr 341

24. Judeo-Arabic+Hebrew, Ox. Bodl. Pococke 295,
Maimonides, Mishnah Commentary

26. Greek papyri (with WÜ, HD, B)

27. Greek papyri (with WÜ, HD, B)

28. eScriptorium used for Dead Sea Scroll Glyph alignment
Automatic letter level alignment
Images of Dead Sea Scrolls by
Shay Halevy Courtesy Israel
Antiquities Authority

30. p. 3558:

31. Please stay tuned for upcoming workshops
Contact: daniel.stoekl@ephe.psl.eu, peter.stokes@ephe.psl.eu
https://escripta.hypotheses.org
Many thanks to
Bibliothèque nationale de France
National Library of Israel (Ktiv!)
Bayerische Staatsbibliothek München
Biblioteca Apostolica Vaticana
Bodleian Library, Oxford
Cambridge University Library
Israel Antiquities Authority, Jerusalem
Staatsbibliothek Berlin, Preußischer Kulturbesitz
Intro tutorial: https://lectaurep.hypotheses.org/documentation/prendre-en-main-escriptorium