🚨Job ALERT🚨! My old postdoc is available! I cannot emphasize enough how much a life-altering position this was for me. It gave me the experience that I needed for my current role. As a postdoc, I was able to define my projects and acquire a lot of new skills as well as refine some I already had.

Excited to be co-editing a special issue of @dhquarterly.bsky.social on Artificial Intelligence for Digital Humanities: Research problems and critical approaches dhq.digitalhumanities.org/news/news.html We're inviting abstracts now - please feel free to reach out with any questions!

Something I've realized over the last couple weeks with finetuning various VLMs is that we just need more data. Unfortunately, that takes a lot of time. That's why I'm returning to my synthetic HTR workflow. This will be packaged now and expanded to work with other low-resource languages. Stay tuned

I've been getting asked training scripts when a new VLM drops. Instead of scripts, I'm going to start updating this new Python package. It's not fancy. It's for full finetunes. This was how I first trained Qwen 2 VL last year.

Let's go! Training LFM2-VL 1.6B on Catmus dataset on @hf.co now. Will start posting some benchmarks on this model soon.

Training on full catmus now and the results after first checkpoint are very promising. Character and massive word-level improvement.

LiquidAI cooked with LFM2-VL. At the risk of sounding like an X AI influencer, don't sleep on this model. I'm finetuning right now on Catmus. A small test over night on only 3k examples is showing remarkable improvement. Training now on 150k samples. I see this as potentially replacing TrOCR.

New super lightweight VLM just dropped from Liquid AI in two flavors: 450M and 1.6B. Both models can work out-of-the-box with medieval Latin at the line level. I'm fine-tuning on Catmus/medieval right now on an h200.

With #IMMARKUS, you can already use popular AI services for image transcription. Now, you can also use them for translation! Transcribe a historic source, select the annotation—and translate it with a click.

GLM-4.5V with line-level transcription of medieval Latin in Caroline Miniscule. Inference was run through @hf.co Inferencevia Novita.

Qwen 3-4B Thinking finetune nearly ready to share. It can convert unstructured natural language, non-linkedart JSON, and HTML into LinkedArt JSON.

Discover Magazine did a nice feature on the Voynich Manuscript. I had a delightful conversation with Sam Waters, and here's the result. (there's also a print version in the current issue) #linguistics #Yale #YaleLibrary #BeineckeLibrary #conlangs www.discovermagazine.com/was-the-worl...

I spent the last 24 hours finetuning Dots.OCR with different datasets. Here are some of the things I learned... TLDR Don't sleep on this model! If you do HTR or OCR try this.

Vibe coding is great for quick visualizers for projects. Single Claude 4 Sonnet prompt and viola.

Finetune of Dots.OCR working! The output isn't quite there yet. Still a lot more training to go, this is just a checkpoint. This is far better than the default model, though! It learns to adjust the bboxes to your style very early in the training, so layout parsing is easier to work into the model.

Woot first got the first finetune of dot.ocr on @hf.co ! Going to share the training script hopefully today. This finetune is for old church slavonic.

Yeah, really impressive. There are some minor errors, but really much better than anything I've tried until now.

@ponteineptique.bsky.social any interest in finetuning this with me on HTR-United data? I'm thinking the logical place is to start with Catmus segmented with the HTR data. It can handle the layout parsing and HTR, so is there a good way to get both data for all catmus?

One of the best HTR/OCR and Layout detection models I have ever seen! And the size is soooo tiny!! @bcgl.bsky.social check this out! Curious about your thoughts for newspaper parsing. Free app: huggingface.co/spaces/Moham...

Hoping to share more soon, but last week I developed a workflow that handles complex relationship extraction from any Wikipedia page of any size. I'll be making this open source soon. We did a few tricks to make it really fast (hint, we rebuilt Wikipedia and Wikidata both locally with LMDB).

Qwen-Image just dropped. Pretty incredible text to image model that can handle long-text! Obviously the output is not great, but this was my first pass at it. There's something here that feels different in this model.

Amazing OCR project

Feature alert: ✂️ #liiive just got a new tool: Photoshop-style Smart Scissors! Automatically trace shape outlines and annotate complex objects in seconds. Try it now at liiive.now #IIIF #ImageAnnotation

I've debugged a few things. Working a lot better now in Browser cache mode, available now on HuggingFace spaces. Free way to use VLMs and object detection models to annotate documents quickly. Integrates with Roboflow.

Working to port VLaMy to an entirely free mode where you can just cache all your data in the browser for a project. Slowly adding all the features from the full version to this user-free version. Available now on @hf.co @danielvanstrien.bsky.social Link: huggingface.co/spaces/wjbma...

Abandoned my e-scriptorium fork and rebuilt a similar application from the ground up with VLMs in mind and leveraging their abilities to not only transcribe documents at the zone or page level, but also describe key parts of the image, like GraphicZones. Hoping to share this finally next week.

Anyone have experience with HTR/OCR in the Batak language? Or does anyone know who I should ask about this?

Just saw these cool notebooks for paleography by Peter Stokes! github.com/pastokes/Com...

Working on something fun for my E-Scriptorium fork =)

This is a really really cool tool for collab IIIF things

I'm using langid in Python for language detection and getting consistently 100% accuracy across 15 languages. Is there something I'm missing? Is it really this good? I'm super skeptical when I see 100% consistently.

Are you at #DH2025 and interested in pre-modern data?! Come by to the session on #openaccess, #medieval #manuscripts and #digitalhumanities. I’ll be talking about ATR/HTR/OCR for #OldNorse. The session starts at 16:00 in Aud B3 and we have AC. :)

Our posting is up for the @jcblibrary.bsky.social Associate Director -- please share! It's a good place to work, imho. 😀 brown.wd5.myworkdayjobs.com/en-US/staff-...

Got it working with HuggingFace and Qwen 2.5 VL now. Simply use your free HF token. Pro users can transcribe around 300 pages a day for free.

Working on something fun =) This lets you create regions (bounding boxes) on any IIIF image and run that through VLMs for transcription, alt-text, etc. This is real time.

Good news! I fine-tuned Qwen 2.5 VL and it now supports Old Church Slavonic. Test it for free in Caracal! The dataset kindly shared by Achim Rabus and contains 7.6k pages. 7k for training, 600 for validation. This is an out-of-scope example from the web. (dataset is private for now)

I try not to link to X, but I don't see this posted here. Pretty interesting arxiv.org/abs/2507.07955 x.com/sukjun_hwang...

Here's an example of this working with medieval manuscript abbreviations. This is multilingual too. You can paste in most modern languages and it will work too. Expect results to degrade a bit as you get further away from the original Latin.

Why is semantic searching relevant for medievalists? Quotes often are changed slightly in texts, making them sometimes difficult to identify. Here's an example. With semantic search, it's possible to resolve variance in words, spelling, word-order, etc. App: huggingface.co/spaces/medie...

Potentially interesting for building models in relatively data-scarce domains (eg historical language models)

Another great open source project from Ben Lee!

#DigitalHumanities Postdoctoral Fellowships

‼️Sentence Transformers v5.0 is out! The biggest update yet introduces Sparse Embedding models, encode methods improvements, Router module for asymmetric models & much more. Sparse + Dense = 🔥 hybrid search performance! Details in 🧵

#DigitalHumanities PhD position

After a lot of hiccups finally finished recording my next series on cuML!

Finishing a series of 7 videos that walks through various tasks, from topic modeling to trend identification to semantic search all on your GPU. Here's a fun graph from the American stories newspapers where a specific topic peaks every November 1938-1945. The cluster? American Football.

New multimodal embedding model from JinaAI has me really excited. This is a very large model, but may be really useful for certain projects. I'll be testing it out with LUX multimodal data in the next week. Model: huggingface.co/jinaai/jina-...

Hoping this can be made public soon. There will soon be a lot of medieval HTR fine-tunes of Qwen 2.5 VL! =) Got a Transkribus dataset that you want to train Qwen 2.5 VL on? Please reach out! More than happy to do the training for you if we can make the model public (pls the data too).

Having fun today. Getting to merge one of my favorite DH projects into LUX at Yale. Building the mapper to map Pleiades places to the LUX Linked-art format via linked open data. Soon we will have equivalents for ancient places thanks to that hard work at Pleiades!

Have high quality data sitting on Transkribus? Want to make it available on @hf.co with a single command line? Introducing Transkribus-HF which allows you to take a Transkribus export zip and make it into a HF dataset! It can parse pages, regions, lines, or windows! github.com/wjbmattingly...