Contribution to the workshop "Des corpus pour l’histoire à l’âge du numérique : Éditions électroniques d’actes royaux et rinciers
(Moyen Âge-première modernité)", 16.3.2018, École Nationale des Chartes, Paris, in the context of the project 'AcRoNavarre' (https://acronavarre.hypotheses.org/)
Scaling up coastal adaptation in Maldives through the NAP process
Digitising charter images : benefits and pitfalls
1. Digitising charter images :
benefits and pitfalls
Georg Vogeler
@gvogeler
http://www.i-d-e.dehttp://informationsmodellierung.uni-graz.at
2. monasterium.net
• 636.658 charters
• 869.240 images:
• 465.099 charters with images
• Ca. 420.000 charter images from
archives
• Ca. 300.000 charter images on
monasterium.net server
http://www.monasterium.net
-
100.000
200.000
300.000
400.000
500.000
600.000
700.000
800.000
900.000
1.000.000
01.02.12
01.06.12
01.10.12
01.02.13
01.06.13
01.10.13
01.02.14
01.06.14
01.10.14
01.02.15
01.06.15
01.10.15
01.02.16
01.06.16
01.10.16
01.02.17
01.06.17
01.10.17
01.02.18
Urkunden
Regesten
Transkriptionen
Bilder
3. A Pluralistic View on Charters
CHI
Intention, Content, Meaning, Semantics
(linguistisc) Code,
System of Symbols,
Iconography, ...
Unique physical object,
materiality
Holistic: object as a
complex sign
CHF
Feature set
Work, Structure
(based on the „Text Wheel“ by Patrick Sahle: Digitale Editionsformen, III, 2013)
4. Transcription
Image
A Pluralistic View on Charters
CHI
Intention, Content, Meaning, Semantics
(linguistisc) Code,
System of Symbols,
Iconography, ...
Unique physical object,
materiality
Holistic: object as a
complex sign
CHF
Feature set
Work, Structure
(based on the „Text Wheel“ by Patrick Sahle: Digitale Editionsformen, III, 2013)
6. Image Creation
• Technology:
• resolution
• colour and size reference
• keep physical intervention as
minimal as possible
• Diplomatics problems:
• charter is more than the main text
• plica
• seals as 3D objects
7. Reflectance Transformation Imaging (RTI)
ImagebyMarkMudge,,http://culturalheritageimaging.org/
Cfr. Franz Fischer / Stephan Makowski: Digitalisierung von Siegeln mittels Reflectance Trans formation Imaging (RTI), in: Paginae historiae. Sborník Národního archivu 25,1 (2017), S. 137-141.
8. Legal issues of digiral photrographies of charters
thanks to Walter Scholger
• It‘s usually not about „copyright“:
• Copyright is an individual right of the creator, ceases 70 years after the
death of the creator
• No „originality“ in operation of standardized procedures.
• Copyright is hold by a natural person – not legal bodies
• It might be about „related rights“:
• Subject to national law with a general frame in the EU InfoSoc-directive
• e.g. substantial investments to create the images and manage them
• It‘s all about „licensing“:
• Find a balance between the interests of researchers and archives, e.g.
• CC-BY-SA? CC-BY-SA-NC? (https://pro.europeana.eu/page/available-rights-statements)
Philipp Maier: Digitization Of Cultural Heritage. CO:OPyright guidelines and recommendations, Graz 2018 (forthcoming)
10. Regions of interest: e.g. Signum Recognitionis
http://monasterium.net/mom/image-collections#Otto%20I.%20Rekognitionszeichen2100132
11. Stiftsarchiv St. Paul, St. Blasien U 391 (= Archiv des
Stifts St. Paul im Lavanttal, Urkunden St. Blasien //
Varia Ecclesiastica BU 429 in monsterium.net)
1226 Dezember , apud tres Sanctos (Xanten)
Kaiser Friedrich II. nimmt die „fratres in ecclisia
Denkendorf dominici sepulchri domino laudbiliter
famulantes“ mit ihren Gütern in seinen Schutz.
1 Siegel (teils gebr.), Perg., Film SIV 689/ 1791
Protoedition? (example: semiautomatic alignment MOM – RI)
RI IV,1,1 n. 1690:
1226 dez. 00 apud Tres Sanctos
Friedrich II. nimmt die brüder des
heiligen grabes in Denkendorf mit
personen und besitzungen in seinen
besondern schutz. Z.: Julian bisch. v.
Mazara, Joh. bisch. v. Boiano, Rich.
marsch. des fürstenthums u. bruder
G. v. Mer(ern) truchsess. Besold Doc.
red. 1,282. Huill. 2,699. Wirtemb.
Urkkb. 3,206. -- [Vgl. letzteren druck
wegen der zeugen; an den truchsess
Gunzelin wird schwerlich gedacht
werden dürfen, es handelt sich wohl
um dieselbe person, welche 1234 als
frater G. de Merk bezeichnet bote
des kaisers beim könige von England
war. Huill. 4,505.]
On protoedition: Olivier Guyotjeannin: Éditions diplomatiques et recherche historique. Quelque remarques sur le cas français (XIXe-XXe siècles), in: Vom Nutzen
des Edierens, hg. v. Brigitte Merta, Andrea Sommerlechner u. Herwig Weigl, Wien 2005 (MIÖG Erg. Bd. 47), S. 303-312
http://monasterium.net/mom/RIViI/1226-12-00_4_0_5_1_1_2433_1690/charter
13. Tobias Hodel: Sending
15th-Century Missives
through Algorithms:
Testing and Evaluating
HTR with
2,200 Documents, IMC
Leeds 2017 Paper, 11th
July 2017
14. Transkribus
• E.g. Hodel with Thun missives (15th century):
• 120 transcriptions (of 2200) as training data
• 20% character error rate without language model, 13% with language model,
but improvement mostly in formulaic texts
Tobias Hodel: Sending 15th-Century Missives through Algorithms: Testing and Evaluating HTR with 2,200 Documents, IMC Leeds 2017 Paper, 11th July 2017
https://solascriptum.wordpress.com/2017/07/11/imc-leeds-paper-sending-15th-century-missives-through-algorithms-testing-and-evaluating-htr-with-2200-documents/
15. Illuminated Charters
• Martin Roland, Andreas Zajic, Gabriele Bartz, Markus Gneiß, Martina
Bürgermeister, Georg Vogeler, FWF funded (P 26.706)
• Classification of decorations (Martin Roland 2014):
• 1: Historiated, with colors
• 2: unconventionally rich
• 3: Graphical signs as means of authentication
• http://monasterium.net/mom/IlluminierteUrkunden/collection
• http://illuminierteurkunden.uni-graz.at
http://illuminierteurkunden.uni-graz.at
18. Illuminated Charters: Confusion Matrix
(Christlein 2018)
Humanattribution
Automatic attribution
Niveau 1 Niveau 2 Niveau 3 Niveau 4 precision
Niveau 1 336 87 4 25 74%
Niveau 2 97 1459 90 288 75%
Niveau 3 1 56 342 75 72%
Niveau 4 32 897 113 6094 85%
Niveau 1 historiated
decoratedNiveau 2
Niveau 3 graphical authentication
Niveau 4 no graphical elements
19. Illuminated Charters: Confusion Matrix
(Christlein 2018)
Humanattribution
Automatic attribution
Niveau 1 Niveau 2 Niveau 3 Niveau 4 precision
Niveau 1 336 87 4 25 74%
Niveau 2 97 1459 90 288 75%
Niveau 3 1 56 342 75 72%
Niveau 4 32 897 113 6094 85%
Niveau 1 historiated
decoratedNiveau 2
Niveau 3 graphical authentication
Niveau 4 go graphical elements
20. What does the machine see?
Niveau 1 Niveau 4
Human: Niveau 2
Charters with drawn (graphic) decoration or display
scripts with decorative character, exceeding the
contemporary standards and/or characteristic for the
production of chancelleries. The graphic decoration
became more and more elaborate from 13th century
onwards. Therefore the selection must refer to a
standard of the specific period.
21. Benefits of Digital Imaging
CHI
Intention, Content, Meaning, Semantics
(linguistisc) Code,
System of Symbols,
Iconography, ...
Unique physical object,
materiality
Holistic: object as a
complex sign
CHF
Feature set
Work, Structure
(based on the „Text Wheel“ by Patrick Sahle: Digitale Editionsformen, III, 2013)
22. Digital Images of Charters
Benefits
• Publish charters
• Make comparative work easier
• Gives the machine access to
further facets of the charter
Pitfalls
• Give the machine only partially
access to the image
• Misunderstand how the machine
process the image
http://www.i-d-e.dehttp://informationsmodellierung.uni-graz.at
You are warned!
by
Georg Vogeler
georg.vogeler@uni-graz.at