Sign in

Digital Medievalist

@digitalmedievalist.bsky.social
3.1K followers 522 following 358 posts

International web-based community for medievalists working with digital media.

PostsRepliesMedia
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
🔗 Sources (2/2) Submit a piece to The Digital Orientalist: tinyurl.com/3zf6n99s
tinyurl.com
Submissions
At The Digital Orientalist, we value diverse voices and perspectives in the field of digital humanities (DH). While we usually invite applications for contributors, editors, and guest contributors,…
020
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
🔗 Sources (1/2) Griffiths (abbreviations): tinyurl.com/3n52x2pk Eveleigh & al. (TEI): tinyurl.com/wweksdpa Pasciuto (semantic web): tinyurl.com/bd464zuk Sawy (Coptic): tinyurl.com/42c8m65z Wadhwa (Indian archives): tinyurl.com/mwcahkky
tinyurl.com
Transkribus in Practice: Abbreviations
In my last article, I wrote about my experiments to improve the accuracy of a Handwritten Text Recognition (HTR) model using Transkribus. This is part of my ongoing work with the ERC-funded project…
110
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
One last thing before we wrap up the week. The DO accepts guest submissions on a rolling basis, so if you have something to share, read more bellow. 👇 Mohamed Ilyes Mechentel (@mohamed-mechentel.bsky.social) from The Digital Orientalist @digitalorientalist.bsky.social
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
The lesson I would pass on to medievalists is this: in a text-recognition pipeline, the most consequential choices are made before any model is trained. The convention is the founding act, and writing it down is itself part of the scholarship.
110
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
Our own response was to align with an existing, semi-normalised standard (OpenITI / Dabīrān), to preserve alternative readings instead of forcing a single one, and to freeze the HTR test set only once the convention had stabilised. We document it and version it, like any other scholarly decision.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
None of this is unique to my project; conventions are a running theme at The DO. Rachael Griffiths shows it in HTR for Tibetan, and it recurs well beyond OCR: TEI and cataloguing, standards for selmantic-web networks, metadata for Coptic resources, and interoperability in Indian-language archives.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
Setting it means a run of concrete decisions: keep optional vowel pointing or not, expand scribal abbreviations or leave them, where marginal insertions join the line, one convention for Arabic and Ottoman Turkish or two. Each choice shapes what later analysis can see.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
The other corpus I work on, the Jerusalem court registers, the sijillāt, has no machine-readable text at all. There the convention is not inherited; we have to set it ourselves. That turns out to be the founding act of the whole project rather than a technical detail.
Screenshot of eScriptorium showing a two-page opening of Sijill 173 with automatically detected text lines drawn across the Arabic script.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
This mattered to us because our two corpora sit at opposite ends. The Istanbul fatwā compilations have already been transliterated in printed editions, so we inherit someone else's convention, much as a medievalist works from a critical edition.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
The comparison also surfaces real history. Two fatwā questions with near-identical wording drew opposite rulings a century apart, over whether a husband denying paternity must undergo li'ān. One jurist ruled that he must, the later one that he need not. It found that pair among tens of thousands.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
The point generalises, and it should interest medievalists as much as us. Whatever you run downstream, whether similarity, distant reading, or named-entity recognition, still sits on top of how the text was transcribed. The convention you settled on decides, in advance, what you can detect.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
The explanation turned out to be simple: the two printed editions had transliterated the Ottoman Turkish differently. Our method was comparing transcriptions, not the underlying texts. That missing 0.18 was the measured weight of a single editorial convention.
100
Digital Medievalist @digitalmedievalist.bsky.social · 09/10/2026
👋 Still the DO here for the takeover ! Here is a small puzzle from my work. We ran a text-similarity test on two Ottoman fatwā collections. Word for word, two copies of the same ruling should have scored 1.0, but came out at 0.82. The algorithm was fine, so the missing 0.18 needed explaining. 🧵
120
Reposted by Digital Medievalist
The Digital Orientalist @digitalorientalist.bsky.social · 08/10/2026
Today's 🧵 is all about 🐉📜!
062
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
🔗 Sources Nayebossadrian (WWQI): tinyurl.com/3r2mtcf8 Wadhwa (Sindhi archive): tinyurl.com/d2a85j3f Woltermann (Indonesian): tinyurl.com/46wxn9uf Mohamed Ilyes Mechentel from The Digital Orientalist @digitalorientalist.bsky.social
tinyurl.com
De-Silencing History: From Private Papers to Public Data in the Women’s Worlds in Qajar Iran Project
This is a post by Zhaleh Nayebossadrian | Sapienza University of Rome Introduction: The Digital Turn in Persian Persian Studies, traditionally grounded in the philological analysis of manuscripts a…
031
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
That is the first entry in the bestiary. Tell us which creature we should set side by side next; suggestions are welcome below.👇
120
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
The possibilities offered through virtually consolidating materials are also explored in The DO e.g. Zhaleh Nayebossadrian’s piece on the Women’s Worlds in Qajar Iran project, Soni Wadhwa’s archive of Sindhi language advertisements, and Leon Woltermann’s essay on Indonesian short stories.
120
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
All three pages sit in the BnF and are digitised in Gallica, yet a medievalist and an islamicist would rarely open them in the same sitting. Only digitisation makes it possible to compare these distant pages at all.
120
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
Across the three pages, the dragon is evil, then a moral warning, then part of the cosmos. No single version is the real one; each tradition turns the creature to its own ends. The dates are worth noting too: a 15th-century French book set beside 18th-century Arabic copies.
120
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
The dragon is not always a threat, though. In this copy of al-Qazwīnī's Wonders of Creation the same creature becomes a constellation, al-Tinnīn, our Draco: a red, scaled serpent coiled across the night sky and read here as a map of the heavens.
A coiled red scaled dragon with a maned, canine-like head, framed by Arabic text, from al-Qazwīnī's Wonders of Creation; it depicts the constellation al-Tinnīn (Draco).
Source: BnF, Arabe 2178, f. 20r
130
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
In the Islamicate world the dragon plays other roles. In this Barlaam and Josaphat it drives a parable: a man flees a beast, climbs a tree above a pit, and a dragon waits below as an image of death. The tale itself had already crossed many cultures to reach this page.
A painting from an Arabic Barlaam and Josaphat: at left a large winged beast with bared fangs; at right a man in a tree above a well, with dragon and serpent heads waiting below.
Source: BnF, Arabe 273, f. 42r
140
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
In the Latin West, the dragon usually stands for evil. In this Bible historiale it appears as the seven-headed beast of the Apocalypse, crowned and monstrous, threatening the radiant Woman while Saint John looks on.
Illuminated miniature: the crowned Woman of the Apocalypse standing on a crescent moon and haloed in gold, two angels carrying a golden ark above, Saint John holding a book at left, and a seven-headed crowned dragon at right.
Source: BnF, Français 4 (Bible historiale), f. 255r
120
Digital Medievalist @digitalmedievalist.bsky.social · 08/10/2026
Hello again, the DO here for the takeover! Many of us work from digitised archives, which let us set side by side manuscripts that never shared a room. As a small experiment, I follow one creature, the dragon, across the Western and Islamicate traditions, in three pages from the BnF. 🐉🧵
182
Reposted by Digital Medievalist
The Digital Orientalist @digitalorientalist.bsky.social · 07/10/2026
✨ Day two of our @digitalmedievalist.bsky.social takeover! Today's 🧵looks at some of the challenges we face when applying digital tools to non-Western manuscripts...
022
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
Irfan (South Asian scripts): tinyurl.com/3ukks4a9 van Lit (RTL Arabic): tinyurl.com/mxbm6cp4 Mohamed Ilyes Mechentel from The Digital Orientalist
tinyurl.com
Scripts That Don’t Fit: The Hidden Bias of NLP in South Asian Languages
This is a guest post by Saniya Irfan. Digital Humanities in South Asia often begins with negotiating invisibility, the invisibility of our languages in NLP resources, datasets, and interfaces. Work…
042
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
The tools are largely the same, then, but they make different demands on us. It is worth noticing the assumptions built into our methods, because that is what lets our two fields learn from one another. More this week. 👇
120
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
None of this makes the sources harder in themselves; it makes them differently shaped, and the borrowing goes both ways. We take up the standards and pipelines that medievalists have built over years, while they can take up our experience of scarce data and many hands on a single page.
120
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
Then there is ground truth, the corpus of transcribed pages a model learns from. For medieval Latin, large shared sets and ready-made models already exist; for Arabic script, far less is available. In practice we often have to build that infrastructure ourselves before we can even use it.
120
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
Take a case from my own work: Sijill 173, a 17th-century Ottoman court register from Jerusalem. It is written in Arabic hands, with the occasional passage of Ottoman Turkish and no modern punctuation. At almost every step it raises a question that a Latin-script pipeline rarely has to face.
A two-page opening of Sijill 173, a 17th-century Ottoman court register from Jerusalem: dense right-to-left Arabic handwriting in dark ink, with no modern punctuation.
120
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
There are also the diacritics (tashkīl), the small marks that fix vowels and meaning. In many manuscripts they are absent, or only partly supplied. So even a basic decision, what counts as the text we transcribe, becomes an editorial and epistemological one rather than a given.
120
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
This mismatch between right-to-left scripts and left-to-right tools has been discussed at The digital Oriantalist: Saniya Irfan on the hidden bias against South Asian scripts, and The DO's founder, Cornelis van Lit, on Arabic in a left-to-right digital world.
120
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
An Arabic-script page behaves differently. The writing runs right to left, the letters join up and change shape according to their position, and the end of a word is not marked in the way a space marks it. The units the software takes for granted, the line and the word, no longer hold.
130
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
Start with how HTR works. The model reads a page a line at a time, so it must first locate those lines, a step called segmentation. It learned this on the Latin alphabet, where the script runs left to right, letters stand in sequence, and spaces divide the words. For Latin, the assumptions hold.
130
Digital Medievalist @digitalmedievalist.bsky.social · 07/10/2026
👋 Hello, this is the Digital Orientalist again, still on our takeover week! We share our digital toolbox with medievalists: same HTR, the same pipelines. What changes, once you turn those tools on non-Western manuscripts, is the method. This thread is about where these approaches meet and part.🧵
1125
Reposted by Digital Medievalist
The Digital Orientalist @digitalorientalist.bsky.social · 06/10/2026
First 🧵 from our takeover!
011
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
🔗 Sources (2/2) Arabic HTR history: tinyurl.com/8x7j3t4h Calfa interview: tinyurl.com/2dfx5hfx Wong (networks): tinyurl.com/9ht68xkb Chow & Iu (Hong Kong): tinyurl.com/yuf3c9cm Mohamed Ilyes Mechentel from The Digital Orientalist
tinyurl.com
At the Dawn of Digital Studies on Arabic Script in France (2) : A Brief History of Handwritten Arabic Text Recognition in France
Introduction The first article of this series explored recent advances in the digital study of Arabic script in France in general. It traced the beginnings of digital studies applied to Arabic scri…
031
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
🔗 Sources (1/2) Morris (Kirishitanban): tinyurl.com/ytr4wewh Ishac (Syriac): tinyurl.com/mthyr3k5 Griffiths (CER): tinyurl.com/y23b2ut6 Kynn (teaching): tinyurl.com/3h45p9me Wion (Ethiopic): tinyurl.com/yx94su7z
tinyurl.com
Transkribus and Kirishitanban: Some Initial Experiments
Inspired by a workshop given by Dr. Alex Mallett (Waseda University) that I attended in May last year, I recently began experimenting with Transkribus,a platform that, amongst other things, allows …
110
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
If one thread runs through all this, it's the pull between open and closed: who holds your data, your models, your conventions. That's why I lean towards open-source, and why the big-AI shortcut worries me. Questions on any of these? Ask below. More this week. 👇
130
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
On Gephi, The DO has two pieces: Brian Tsz Wong @brianthwong.bsky.social on network analysis in the classroom, and Eric Chow and Ryan Iu on navigating early colonial Hong Kong history.
120
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
Beyond HTR: if your work leans on annotation or network analysis rather than transcription, two accessible open tools are worth a look. Recogito, for annotating texts and images (places, people), and Gephi, for visualising networks.
131
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
A shift worth flagging 🚩: Big AI models keep getting stronger on such sources. Ethically it's off. Heritage scraped into closed models, with no consent or benefit to its archives, runs against open, reproducible work. And it's not optimal yet: they confabulate, with no control over ground truth.
131
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
There is no single best. My own preference is eScriptorium. Personally, the technique is not that hard to acquire. For a beginner, though, I'd point to Transkribus: it is easier to start with, and its cooperative model sits better with me than a purely commercial one.
120
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
The trade-off runs opposite to Calfa: Arkindex is open and your data exports cleanly, but running it yourself is heavy (a Linux server, Docker, a database, a GPU). Teklia advises against self-hosting without a seasoned sysadmin, so most projects lean on institutional IT or Teklia.
131
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
4. Arkindex Also open-source, but built for more than transcription: it can organise large corpora, spot named entities, and classify pages. There is more to set up, and it is designed for scale. It is worth knowing once a project grows.
110
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
One caveat: that ease often comes from working with Calfa's team, who can assess your corpus, help annotate, or train the model for you. A dedicated model then stays on their platform, it is not yours to download.
110
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
3. Calfa Vision It is built for "difficult," non-Latin scripts: Armenian, Arabic, Syriac. It needs no code, and it learns well from very few annotated pages, a useful option when you are starting almost from scratch.
110
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
The DO has more on Transkribus than any other tool, so for examples on non-Latin scripts and non-European languages: James Morris on Kirishitanban, Ephrem Ishac on the first public Syriac model, Rachael Griffiths on lowering error rates in Tibetan, and Tyler Kynn on teaching with it.
110
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
2. Transkribus @transkribus.bsky.social Probably the most accessible to begin with: a polished interface and a large catalogue of ready-made models. It is credit-based: a monthly free tier, then paid. It is strongest on European hands. One limit: models cannot leave its ecosystem.
110
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
1. eScriptorium 🆓 I use it daily. It is open-source, browser-based, and strong on right-to-left and non-Latin scripts. There is a catch: you need a server to run it, and shared instances often ask you to release your data and models openly in return.
Screenshot of eScriptorium: an Arabic manuscript page on the left, and on the right a panel of numbered transcription lines being entered.
110
Digital Medievalist @digitalmedievalist.bsky.social · 06/10/2026
One thing worth knowing upfront: these tools learn from ground truth, pages a human has already transcribed, which the model treats as the right answers. The more you give it, the better it reads. A number of programs handle HTR. These are the four that feature most in The DO and my own work.
110