#HTR

2025-06-22

Delighted to announce that the paper on best practices in pre- and post-ATR for historical research that I co-authored with @koen_hufkens is now available on Zenodo:

zenodo.org/records/15703825

Big thanks to @hermes_datenkompetenzzentrum for funding our initial workshop at @DHLab_IEG earlier this year. It was nice to present our findings to a larger, international community at #dae2025 in Göttingen. #histodons #digitalhumanities #ocr #htr

Auden in Austria DigitalAAD_OeAW@fedihum.org
2025-06-18

The model works if it gets the errors right.

@transkribus @ACDHCH_OeAW @oeaw @fwf @altbot

#HTR #Transkribus #AustrianAuden

The image shows a handwritten note on a textured paper background. The note is addressed to "Dr Spiel" and expresses gratitude. The text is written in blue ink and is slightly uneven, suggesting it was written by hand. The note is divided into four lines, with the first line reading "Dear Dr Spiel," the second line "I'm enormously grateful to you," and the third line "for all the trouble you're taking." The fourth line is partially obscured, but it appears to continue the message of gratitude. The handwriting is legible, with some words underlined for emphasis. Below the handwritten note, there is a digital transcription of the note, with the same text and corresponding line numbers. The date "August 23rd" is visible at the bottom left corner of the image.

Provided by @altbot, generated privately and locally using Ovis2-8B

🌱 Energy used: 0.173 Wh

Edit: the message really reads "the trouble your taking" - an error that the auden_05 Transkribus model accurately records.
Adi Keinan-Schoonbaertadi@glammr.us
2025-06-13

#britishlibrary RSE Saira Akhter just blogged with reflections from the IIIF Annual Conference 2025 - including some @universalviewer goodies and #IIIF implemented with #OCR #HTR workflows blogs.bl.uk/digital-scholarshi @BL_DigiSchol

2025-06-12

Neuer Beitrag! Julian Helmchen - Die Digitalisierung der spätmittelalterlichen Grundbücher Wiens mit #Transkribus! #htr #histodons #medievodons #auxhist #dighum mittelalter.hypotheses.org/366

Henrik Schönemannlavaeolus@fedihum.org
2025-04-17

In light of current news coming out of UK, a repost:
@SafeguardingResearch has torrents of #trans (healthcare) related resources on SciOp:
sciop.net/tags/trans

Those are all webarchives, easily offline/locally re-playable via replayweb.page/

Re torrenting:
sciop.net/docs/using/torrentin

Most of them come from a list originally made by @dee; mostly webarchived by @ashley
#ProtectTransLives #HTR

Henrik Schönemannlavaeolus@fedihum.org
2025-04-15

Circling back to the starting point of @SafeguardingResearch, there are now torrents of #trans healthcare related resources on SciOp:
sciop.net/tags/trans

Those are all webarchives, easily offline/locally re-playable via replayweb.page/

Re torrenting:
sciop.net/docs/using/torrentin

Most of them come from a list originally made by @dee; mostly webarchived by @ashley
#ProtectTransLives #HTR

2025-04-11

#AppelàParticipation TranscriboQuest, un transcribathon de formation intensive en #HTR et transcription de textes

🗓️ 3-5 septembre 2025
📍ENS de Lyon, site Descartes

📝 Candidature avant le 30 mai !

👉 transcriboquest.sciencesconf.o

#HTR #transcription #formation #biblissima

Jiji, the catjiji@ohai.social
2025-04-11

i was asked at work about machine learning handwriting recognition tools/apps, preferably something that you can train on your own handwriting.

the use case is digitalising handwritten paper notes, so scanning them in and then transforming it into somewhat formatted text.

i tried apples text recognition and it works well, but still makes a lot of mistakes and doesn’t really do any formatting.

anyone can recommend apps or have any experience with this?

#handwriting #recognition #htr #apple

Annika Rockenberger (she/they)arockenberger@fedihum.org
2025-04-10
David HaskiyaDavidH@glammr.us
2025-04-01

A bit depressing how often university research projects doing HTR are unwilling to share their training data with a public agency who wants to combine it with other training data and create an openly licensed HTR-model. To me it seems contrary to an Open Science-ethos, a waste of effort, and be counter-productive to future research.

(Sorry if this comes off as a gripe.)

#HTR #OpenScience #OpenData

2025-03-31

Heute wieder in Köln für das #VL-Anwendertreffen. Spannende Vorträge im Plan. Von mir gibt’s morgen eine kleine Tour durch Nachlassdigitalisierung, #HTR und #NER (und die Grenzen dessen).

2025-03-21

Ahead of Print in BFP: New article: Digitale Erschließung historischer Bibliotheken. The Bibliotheca Eugeniana Digital project applies digital humanities to explore a major Baroque book collection. #DigitalHumanities #HistoricalBooks #MachineLearning #HTR #Visualization

doi.org/10.1515/bfp-2024-0074

FIZ ISE Research Groupfizise@sigmoid.social
2025-03-19

Very happy to have Heleen Wilbrink and Annelot Vijn from Utrecht Archives in our (online) Journal Club today taking about recent projects related to Records in Context (RiC-O) and Persons in Context ontologies as well as on transkribus-based HTR, LLM-based post processing and information extraction.

#archives @epoz @tabea @AnnaJacyszyn @heikef @enorouzi @joerg @sourisnumerique @GenAsefa #ontologies #semanticweb #lod #dh #culturalheritage #htr #transkribus #llms #journalclub

Screenshot of our Zoom-Session of the FIZ ISE Journal Club featuring Heleen Wilbrink and Annelot Vijn from Utrecht Archives.
Handschriftenportalhsprtl@openbiblio.social
2025-03-18

📣 #CfA: Bis 23. April für den zweitägigen Cross-NFDI-Graduiertenworkshop 'Schriftkulturen des Mittelalters in digitaler Perspektive' bewerben! Impulsvorträge & Hands-on-Übungen zu Themen wie #forschungsdatenmanagement, #HTR, digitale Plattformen & #Materialanalyse.

22./23. August 2025 an der @stabi_berlin. Alle Informationen 👉 sbb.berlin/29bt8

#manuscripts #digitalhumanities #medievalsky #dh #handschriften #summerschool

Ankündigungsgrafik des Workshops. Man sieht im Hintergrund ein Foto einer Handschrift sowie einen Computer im Hintergrund. Ein Vordergrund ein blauer Banner mit dem Text "Schriftkulturen des Mittelalters in digitaler Perspektive"
FIZ ISE Research Groupfizise@sigmoid.social
2025-03-14

This week, our colleague @epoz@blog.epoz.org was attending a digitization project end conference at Utrecht Archief with lots of Linked Data presentations

@epoz@mastodon.nl #digitalization #archives #utrecht @utrechtuniversity @fiz_karlsruhe #linkeddata #lod #knowledgegraphs #htr #ocr #transkribus

Photo taken from the back of the audience at the project end meeting at Utrecht Archive.
Auden in Austria DigitalAAD_OeAW@fedihum.org
2025-03-11
A tan piece of paper with the word "Freigesphrochen !" written in blue ink is encased in a light blue jagged line. A thin purple line is visible through the middle of the word. The word is also written in black text beneath the paper.

Provided by @altbot, generated using Gemini
2025-03-09

Interessanter Beitrag im aktuellen Heft ARCHIV. theorie & praxis (ab Juni online) über "Das Digitalisat als Datenquelle. Ein Workflow für die Nutzung bereits digitalisierter Archivbestände".
PERO OCR kannte ich noch nicht.
#ATR #HTR
archive.nrw.de/landesarchiv-nr

2025-03-06

I am very excited to share that I started my PostDoc at Stockholm University this week!

Looking forward to working on layout analysis and handwritten text recognition for rare and unknown scripts, as part of the recently started DESCRYPT project, led by Beáta Megyesi!

#StockholmUniversity #DESCRYPT #htr

Auden in Austria DigitalAAD_OeAW@fedihum.org
2025-03-04

auden_05 #Transkribus #HTR model for the handwriting in pen (English and German) of British-American poet #WHAuden in the period 1957-1973.

We're getting there. 💪

@ACDHCH_OeAW @oeaw @fwf @transkribus @altbot

#AustrianAuden #DH #DigitalEdition #SDE #DSE

The image is a graph titled "Learning Curve". The Y-axis is labeled "CER" and ranges from 0% to 100%. The X-axis is labeled "Epochs" and ranges from 0 to 260. There are two lines on the graph: a blue line labeled "CER Train" and a red line labeled "CER Validation". Both lines start near 100% and decrease rapidly before leveling off. The blue line levels off at a lower value than the red line. A solid green vertical line intersects the right side of the graph. Below the graph are the values "CER on Train Set: 0.83%" and "CER on Validation Set: 5.04%".

Provided by @altbot, generated using Gemini
David HaskiyaDavidH@glammr.us
2025-02-19

We've launched an updated version of our HTRFlow demo-app on HuggingFace.

huggingface.co/spaces/Riksarki

Try it out and give us your feedback!

There are lots of architectural updates under the hood, making it easier for us to maintain the app and add new capabilities like support for Pylaia-models. It's now basically a client to our HTR/OCR-pipeline software HTRFlow Core.

#HTR #ML #OpenSource #DigitalHumanities #Archives

Client Info

Server: https://mastodon.social
Version: 2025.04
Repository: https://github.com/cyevgeniy/lmst