davidlassner

@oli_0331 I've installed it with docker and it was really only those four/five commands as in the instructions gitlab.com/scripta/escrip…

Our Humanistic NLP dream team from left to right: @davidlassner @apjanco, moi, Natasha Ermolaev (locked out of Twitter — apparently “a long story” 😀) & @thatbudakguy. We loved presenting our @DARIAHeu + @PrincetonDH collab at the #DH2023 Language Models sessions earlier today.

Now online 😍 ! My book "From Handwriting to Footprinting: Text and Heritage in the Age of Climate Crisis" is now published and can be downloaded at: openbookpublishers.com/books/10.11647…

Woooo! 😳 Our #DraCor team just won the 2022 #Rahtz Prize for #TEI Ingenuity. What! An! Honor! Thanks to @TEIconsortium – and thanks to all the "DraCorians" out there.

📢to all historians looking for digital tools to analyze historical sources: CorDeep, a #ML based web application, is able to extract 4 different classes of visual elements within historical sources from the 15th to the 17th centuries. 👉bifold.berlin/news-events/ne…

I am delighted to share that "Domain-Specific Word Embeddings with Structure Prediction" has been accepted to TACL. This is joint work with @davidlassner @AnneBaillot & Shin Nakajima @coastalcph & @bifoldberlin. Preprint here: arxiv.org/abs/2210.04962. Code follows soon! TL;DR:

📝 Preprint is now online at arxiv.org/abs/2210.04963 👩‍💻 Code here: github.com/stephaniebrand…

@deliprao @github Currently, I think it is easy to set up code-templates to guide Copilot into wrong directions. If the task is: 'write a `get_index(ll,v)` function' (no assumptions on `ll`) and you put `from bisect import bisect_left` at the top of the template, Copilot will assume ll is sorted.

Copilot-proof assignments™️ are going to be a thing for CS educators. I’m designing assignments for a future course, and I’m assuming students will use @github’s Copilot to solve them. That should be okay and even encouraged. So what are Copilot-proof assignments?

it's here zenodo.org/record/6891852

Just saw that @stjoweil uploaded a kraken model for German fraktur. I just want to thank you! Are compatible with eScriptorium and have a very solid recognition without any fine tuning on my documents. It will be a jump start for fraktur projects that rely on a free OCR pipeline.

@quinnanya Ermordet von Weibo, stilsicherer nickname, aber cancel culture? Maybe it improves if you don't lower-case everything? But it is probably all lower-case in the sources?

@ttasovac I love it, which implementation is most stable? Is there one that supports translations in both directions out of the box?

I like the idea behind the recently published spec for Invisible XML. You write a grammar to describe your plain-text data and the ixml processor automagically turns it into xml. It’s like Markdown on crack. Could be a cool first step in many xml pipelines invisiblexml.org/1.0/

The Sphere project is an incubator of multiple #Digital #Humanities (DH) approaches. Want to learn more about the evolution and transmission of knowledge in the early modern period? 👉bifold.berlin/homogenization…

We invite the TEI community to nominate prospective candidates for the Rahtz Prize for TEI Ingenuity 2022: rb.gy/ynxxk4. Nominations and self-submissions should only be submitted through this form: rb.gy/xcmoe6. Deadline: 30 June 2022 by midnight (HAST).

@AnneBaillot 😍

yes yes yes yes yes

May 11–12: Join us for the New Languages for NLP: Building Linguistic Diversity in the Digital Humanities Conference feat. participants from the @nlp_new Institute and keynotes by David Bamman and Ines Montani. In person and online. Open to all. Sign up! cdh.princeton.edu/updates/2022/0…

@AnneBaillot Actually, I wouldn't be surprised if you had translated books named Shakespeare from a French author into English but the publication date threw me off

😂You never know what I'm up to while you're not looking (yes, I am a translator ghost, too!)

That moment where you look at the results of last night's model training on a new data set and, in shock, realise you must have screwed up your data preparation - why is my PhD supervisor a label in my dataset? @AnneBaillot (turns out gutenberg.org/files/53490/53…)

@cneudecker I use copilot for drafting docstrings and for smaller functions with speaking names it is quite good

Where is the next AI model that can infer readable and comprehensible documentation from raw source code?

@quinnanya +1 for postgres + datagrip