Johannes B. Gruber

@jbgruber.bsky.social

Senior Researcher @gesis.org // Data Editor @polcommjournal.bsky.social 🔎 political communication (#polsky + #commsky) with text analysis and #rstats (#opendata + #openscience) 🌏 JohannesBGruber.eu 👨‍💻 research software github.com/JBGruber

So the "That's not a knoife, THIS is a knoife" country (🇦🇺) confiscated my nail clippers and a metal credit card with a sharp edge, while Christchurch Airport security took a look at my accidentally hand-luggaged Swiss Army knife and went: "Yeah, nah, she'll be right, mate. Blade's within limits." 😅

🔐 MEDem AmCAT tackles a different problem: much of the news and platform content we study cannot be openly shared because of copyright, privacy or licensing rules. AmCAT supports controlled access and analysis without distributing protected material.

Figure 3. Non‐consumptive analysis of a protected corpus in MEDem AmCAT: Instead of giving researchers access to the data, the analysis models come to the data. Notes: An external researcher can submit a prompt for a large language model or a supervised classifier trained on data they do have access to; inside the MEDem infrastructure (dashed boundary), the models read the protected full texts and write their annotations back to the corpus as new shareable fields; only metadata and derived data cleared for sharing are returned to the researcher; continuing the example from Figure 2, the view is filtered to the documents annotated as political (IDs 1 and 3), which are now further annotated by topic (large language model) and tone (classifier).

I needed to come to Brisbane to see people being good at riding e-scooters. I wasn't even aware such a thing existed. But quite often I see people in groups casually doing a wheelie here. And just now, someone drifted around a corner at full speed on a lime. Pretty wild.

New #rstats 📦: mediacloud2 (happy about name recommendations). Since the exitsting R packages were stale and I needed data, I wrote a new package. MediaCloud is an amazing resource for getting news media data, especially if you rehydrate the articles (e.g., through paperboy).

MediaCloud API (v4) Wrapper

A simple wrapper for version 4 of the MediaCloud API, which provides programmatic access to the MediaCloud archive of online news content (<https://www.mediacloud.org>).

jbgruber.github.io

Right after my last post I see this. Absolutely. I also hate how AI is shoved in my face constantly. There are a few tasks at which it is great, and a million where I don't need or want AI and I especially hate that everything is called Gemini or Copilot now.

Tom Scocca@tomscocca.bsky.social · 2mo ago

It's OK at doing code for you? Great. What does that have to do with how it keeps popping up to demand to rewrite a paragraph that's already written and edited?

Qwen3.8 27B comes extremely close to Claude in programming tasks. It's been said, but I wanted to share a test I ran in #rstats for {paperboy} 📦. TLDR: Qwen in opencode got it done in 28 mins, Claude Code ( Opus 5) got it done in 3.5. Results differ a little, but both are perfect solutions.

Bild

While we are still waiting for the final funding decision (which will be made by the BMFTR) for RIDLOP, we are delighted to see the Wissenschaftsrat's report that was published yesterday. So keeping our fingers crossed for the last steps in the evaluation process!

GESIS - Leibniz-Institut für Sozialwissenschaften@gesis.org · 4mo ago

#RIDLOP #Wissenschaftsrat #ResearchInfrastructure The German Science and Humanities Council (Wissenschaftsrat) supports the implementation of RIDLOP, the Research Infrastructure for Data from Large Online Platforms, in its statement published today.