Welcome to the eScripta blog. This is a ‘work in progress’ blog associated with eScripta, the Digital Humanities team within the Scripta project at Université Paris Sciences et Lettres. It will feature regular discussions from members of the project team and selected external contributions. Articles will comprise a mixture of presentations of work in progress, discussions of methodological challenges, and notices of future events and related news. The target audience is those in Computational and Digital Humanities and (Historical) Document Analysis, including related specialisations such as Computational and Digital Palaeography (modelling and describing script, OCR and HTR, layout analysis, and so on).
Almanach and the eScripta team are very happy to welcome Gernot Fink, professor of pattern recognition at the Technical University of Dortmund, and Eugen Rusakov for a workshop on “deep learning for word spotting”.
INRIA, Centre de Paris, 2 Rue Simone IFF, 75012 Paris, Salle de Reunion Centre C434
- 9-10:30 theoretical introduction (G. Fink)
- 11-12:30 introduction (Part 2)
- 14-16 hands-on workshop (E. Rusakov)
- 16-18:30 workshop (Part 2)
Participation in the hands-on workshop is limited to 10 persons. Please contact firstname.lastname@example.org before Tuesday, Oct 15, 2019.
Le 17.10.2019 Almanach et l’équipe eScripta ont le plaisir d’accueillir Gernot Fink, professeur à l’Université technique de Dortmund, et Eugen Rusakov pour un workshop (en anglais) dédié au “deep learning for word spotting”.
INRIA, Centre de Paris, 2 Rue Simone IFF, 75012 Paris, Salle de Reunion Centre C434
- 9-10:30 introduction théorique (G. Fink)
- 11-12:30 introduction (2ème partie)
- 14:00-16:00 workshop hands-on (E. Rusakov)
- 16:30-18:30 workshop (2ème partie)
La participation à la deuxième partie (pratique) est limité à 10 personnes. Merci de contacter email@example.com avant mardi 15.10.2019
Relevant recent publications:
- Attribute CNNs for Word Spotting in Handwritten Documents. Sebastian Sudholt and Gernot A. Fink. Int. Journal on Document Analysis and Recognition, 21(3), pages 159-160, 2018. [bib] [pdf] [http] [view]
- Learning Deep Representations for Word Spotting Under Weak Supervision. Neha Gurjar, Sebastian Sudholt and Gernot A. Fink. Proc. Int. Workshop on Document Analysis Systems, Vienna, Austria, 2018. [bib] [pdf]
- A Probabilistic Retrieval Model for Word Spotting Based on Direct Attribute Prediction. Eugen Rusakov, Leonard Rothacker, Hynuho Mo and Gernot A. Fink. Proc. Int. Conf. on Frontiers in Handwriting Recognition, Niagara Falls, USA, 2018, Winner of the IAPR Best Student Paper Award. [bib] [pdf]
- Gernot A. Fink, Deep Learning for Word SpottingTutorial presented at Int. Conf. on Frontiers in Handwriting Recognition (ICFHR), Niagara Falls, USA, 2018. [bib] [pdf]
The eScripta team has recently returned from the International Conference of Document Analysis and Recognition (ICDAR) which was held this year in Sydney. As well as the main conference, there were also very interesting and useful pre-conference workshops, including HIP (the 5th International Workshop on Historical Document Imaging) and ICDAR-OST (the 2nd International Workshop on Open Services and Tools for Document Analysis). The eScripta team gave two papers:
- BADAM: A Public Dataset for Baseline Detection in Arabic-script Manuscripts: B. Kiessling, D. Stökl Ben Ezra and M. Thomas Miller (HIP)
- eScriptorium: An Open-Source Platform for Historical Document Analysis: B. Kiessling, R. Tissot, P. Stokes and D. Stökl Ben Ezra (ICDAR-OST).
Both papers were very well received, and we also had a lot of interesting discussions with other researchers in the field. We hope some new collaborations will come out of this, and certainly we have many new ideas, so watch this space to see what comes out of it all!
We are happy to announce that the European Research Council has accepted our INFRADEV project proposal “Resilience”. Please stay tuned for further news.
As mentioned in a previous post, eScripta is represented at Digital Humanities 2019 in several different sessions. We now have the full details:
- What: A workshop on Archetype took place yesterday, in which participants installed the system on their computer and learned how to apply it to their own research.
When: Workshop WS-07, Monday (9 July), all day
- What: I (Peter Stokes) will be speaking on a panel about ERC funding.
When: Thursday lunch (11 July), 12h30–13h30
Where: Pandora Foyer
- What: There will be a poster on eScripta, and Ben Kiessling and I (Peter Stokes) will be there to talk to you about it and give live demonstrations.
When: PS Poster Session, Thursday afternoon (11 July), 15h30–17h
Where: Plein 6
- What: Ben will be giving a short paper on ‘Kraken, A Universal Text Recogniser for the Humanities’.
When: Session SP-26, Friday morning (12 July), 11h–12h30
Where: Cloud 9
We very much look forward to seeing you there!
On the 4th and 5th of June in London, the French Embassy in the U.K. and King’s College London are jointly funding a workshop to bring together domain experts from London and Paris to explore the difficulties surrounding governance, funding, and sustainability. The workshop will be coordinated by King’s Digital Lab and discussions will be led by researchers from KCL and PSL, including Peter Stokes, Ben Kiessling and Robin Tissot from the eScripta team. Here is the description of the workshop from the King’s Digital Lab site:
Analysis of hand written materials is by nature intensive and exacting work. Accordingly this topic is a tempting target for the application different digital approaches. Regardless, the joint challenges of unifying the study of multiple scripts, automating classification, and allowing for a workflow that incorporates the indispensable eye of the seasoned palaeographer remain worthy (and elusive) goals.
Beyond the technical hurdles of effective software development there is a minefield of sustainability questions around maintaining the stability, security, availability and continued relevance of any promising digital tool. Engaging and developing a supportive community can help to ease the burden on developers and institutions but the social platforms needed to build these communities require their own maintenance. …
In addition tackling these issues we aim to cement and build upon existing international research relationships in a collaborative spirit. These two focused days will produce new insight into these questions that will be relevant to Research Software Engineering beyond its application in the study of written communucation.https://www.kdl.kcl.ac.uk/events/written-cultural-heritage-written-cultural-heritage-advanced-methods-workflows-and-software-development-approaches-digital-humanities-research/
This joint discussion is particularly important because we expect that Archetype (currently maintained by the King’s Digital Lab) will form one of the key blocks in eScriptorium, but it will also be an important opportunity to share experiences, ideas and practices in the two different countries and institutions. We at eScripta are grateful to King’s and the Embassy for this support, and we are looking forward to the workshop and the future collaborations that it might bring.
Our platform for automatic manuscript transcription will be presented in the talk “eScripta: An Open Source Platform for Artificial Intelligence and Deep Annotation Applied to Written Objects” by Daniel Stökl Ben Ezra at the Institute for the Study of the Ancient World, New York on Thursday, April 4, 2019 between 12h30-1h30.
Address: 15 East 84th Street, New York, NY 10025.
In the frame of the research programme Scripta at Université PSL in Paris, we are developing eScripta, an open source virtual research environment for analyzing and publishing manuscripts and inscriptions. It combines combining quantitative approaches from AI such as automatic layout analysis and automatic transcription of manuscripts (HTR / OCR) or automatic textual criticism and stemmatology with qualitative approaches, deep paleographical, philological, linguistic or historical annotation, in bricks such as kraken by Benjamin Kiessling, Archetype by Peter Stokes or pyrrha by the ENC. We will present the current state of the platform as well as preliminary research results from projects linked to it dealing with automatic transcription of Medieval Hebrew manuscripts (sofer mahir with U Maryland, Dicta and the National Library of Israel) and crowdsourcing their corrections (tikkoun-sofrim – with Haifa U and the National Library of Israel) where we have reached character error ratios of up to 2% for Latin and Hebrew Medieval manuscripts.
Although not strictly related to eScripta, readers of this blog may still be interested to learn of a series of practical and theoretical workshops I will be giving in April 2019 on paleografia digitale at the Università degli Studi di Firenze, Dipartimento SAGAS. The workshops will be held in Italian, and the deadline for registration is 5 April. Details are in the programme below (click on the image to download the PDF).
Our upcoming open source infrastructure eScriptorium for automatic manuscript transcription and deep annotation has been presented in the framework of the daily kraken workshops by Benjamin Kiessling and Robin Tissot with the support of Peter Stokes and Daniel Stoekl at the spring school manuSciences 19 in Fréjus at the Côte Azur, March 10-15, 2019. We met quite some enthusiastic testers.
Have a peek look here
eScripta has been invited to be presented at the Center for Judaic Studies at Princeton University on April 5, 2019 at 12 pm in a talk entitled “Artificial Intelligence, Citizen Science and Medieval Hebrew Manuscripts“.
The Scripta project and eScripta team will be represented at the International Medieval Congress in Leeds this year. Scripta is co-sponsoring a strand of three sessions on ‘Materiality in Series’, and two members of Scripta will be giving papers as part of a strand sponsored by Digital Medievalist. Details are given below. Of course these represent just a very small fraction of the many interesting papers and strands at the IMC, so we very much hope to see you at what looks like being some extremely productive discussions.
Materiality in Series I: Serial Manuscript Production
Thursday 4 July, 9:00 – 10:30 (Session 1526)
New Insights into Serial Manuscript Production in the Early Middle Ages (Tino Licht, Heidelberg)
- Individuality in the Decor of Manuscripts Produced in Series (Lisa Horstmann, Heidelberg)
- Manuscript Production of Verse Romances in French: Re-Grouping Objects, Re-Thinking Texts (Pietro Andrea Martina, Zürich)
Materiality in Series, II: Text Series in Manuscripts
Thursday 4 July, 11:15 – 12:45 (Session 1626)
- The Lives of Saints in Series: A Perspective from Old French Manuscripts (Ariane Pinche, Paris)
- A Reasoned Series of Texts?: The 9th and the 10th Fascicule of the Manuscript Firenze, Biblioteca Nazionale, MS Banco Rari 217 (Nicolò Premi, Paris)
- Songs in Series: Order and Distribution of Peirol’s Compositions in Old Occitan Sources (Stefano Milonia, Rome)
Materiality in Series, III: Serial Analysis and Materiality
Thursday 4 July, 14:15 – 15:45 (Session 1626)
- Can We Explore the Materiality of Charters Using Digital Methods? (Nicolas Perreaux, Frankfurt & Paris)
- Serial Analysis of Manuscript Features: Placing Epic Manuscripts inside French Vernacular Production (Jean-Baptiste Camps, Paris)
- Serial Analysis: The Materiality and Texts of Relic Labels (Kirsten Wallenwein, Heidelberg)
Digital Materiality, II: How to Represent Materiality Digitally in Palaeography and Codicology? (Sponsored by Digital Medievalist)
Monday 1 July, 16:30 – 18:00 (Session 324)
- Peter A. Stokes (eScripta, Paris): Modelling Writing: Towards a Conceptual Reference Model for Palaeography
- Marc H. Smith (Scripta, Paris): On the Epistemological Limits of Automatic Classification of Scripts
- Carolin Schreiber (Munich): Book Covers as Material Objects: Possibilities and Challenges in the Brave New Digital World
The eScripta team is very happy to announce that we will be presenting at DH2019, the annual Digital Humanities conference this year in Utrecht on 9–12 July. We will be presenting a poster on the eScriptorium platform that is currently under development and that focusses on a web interface to Kraken, Archetype, and (most likely) other components such as Pyrrha (from the École des Chartes) and TEI Publisher.
Peter Stokes from eScripta will also be giving a full-day workshop with Debora Matos from the University of Münster, to show the theory and practice of using Archetype for your own research into palaeography, art history and other studies of written communication.