Joan Giner

Improving the world through improving data! It is such an honor to be part of this project's responsible AI team. Here is the Responsible AI spec: mlcommons.github.io/croissant/docs… We need hands! Participate at github.com/mlcommons/croi… @UOCuniversidad @LIST_Luxembourg @JordiCabot

🔴 Cap a la manifestació més gran de la història per Palestina a #Barcelona ‼️ ➡️Resistint al colonialisme, al genocidi, a l'apartheid i a l'ocupació. ➡️Fi al comerç d'armes i a les relacions amb Israel. 🗓️ 4 d'octubre, 12h, Jardinets de Gràcia, BCN

🎉 This Oct at #IDW2025 (Barcelona Supercomputing Center) we will present our AI-Ready Data Workflows for Social Sciences & Humanities: 1️⃣ Lower entry costs with public HPC 2️⃣ Better data curation & AI validation 3️⃣ Reproducibility via RO-Crate-like solutions 👉 Let’s connect!

🎉 Excited to share our accepted paper! We explore how Model-Driven Engineering (MDE) can boost AI/ML by: ✨ Integrating data-centric practices 📂 Enabling open, domain-specific datasets ⚖️ Embedding ethical principles into the lifecycle Read more 👉 dl.acm.org/doi/10.1145/37…

👉 Avui, la lluita per la ciència és també lluita antifeixista. ✊📚🔬ara.cat/internacional/…

La situación en la Franja de Gaza es cada vez más preocupante y dolorosa. Renuevo mi ferviente llamamiento para que se permita la entrada de ayuda humanitaria digna y se ponga fin a las hostilidades, cuyo precio desgarrador pagan los niños, los ancianos, las personas enfermas.

⚠️ Ante el #apagón de la red eléctrica recordamos que las personas trabajadoras cuentan con permisos retribuidos en caso de no poder desplazarse o desempeñar su trabajo. ‼️ Ante esta situación pedimos mantener la calma y seguir las instrucciones de las autoridades.

@Dorialexander @Laz4rz @MartaVillegasM We are currently doing all the HTR workflow using Kraken, and a set of manuscripts manually transcripted by experts. Any recommendations/tool to improve the workflow will be very valuable! And yes! Our goal is to publish all the manuscripts + transcriptions!

@giner_joan @Laz4rz @MartaVillegasM Well ancient Catalan manuscripts can also be within the scope of Commnon corpus :) (was just having a meeting with an IBM team over OCR/PDF parsing)

Why choose. (Mare Nostrum in Barcelona: yes it's a church)

@Laz4rz @Dorialexander Definetly! please share the results as you got them. Putting @laiacastroh in this thread, as she leads some related research lines in this topic.

@giner_joan @Dorialexander That’s very interesting. We have a hanging experiment in which we use a simulacra like experiment to assess election results based on big5 personality traits and real life data — sounds like it would just in your ballpark.

@Laz4rz @Dorialexander I will read it, thanks!!. We also have ongoing works in social science analyzing social networks for political ecology, political polarization, and so on! We will release the topic we are working on very early (we are a very new group at BSC). And yes! If you visit BCN, write!

@giner_joan @Dorialexander Holy cow, would it also apply to me? Don’t know what exactly you guys do with digital humanities (will do in the evening), but we have also done some work in this area and maybe it’s something you’ll find interesting: arxiv.org/abs/2412.16772

@Dorialexander @Laz4rz The language unit led by @MartaVillegasM should be the team at BSC that has used the Common Corpus to train their models. At the SS and Humanities team, we are training models to understand ancient Catalan Manuscripts. Anyway, just ping me, and we can arrange a meeting here :)

@giner_joan @Laz4rz We should definitely connect more. I’ve noticed you’ve trained on some part of Common Corpus lately and we have better versions of the data internally.

@Dorialexander @Laz4rz If you come to Barcelona and want to see it, ping me! 🤗 We are also doing cool stuff in digital humanities and AI that I hope you will find interesting!

@Laz4rz Haven't been there yet, but there is a virtual visit bsc.es/discover-bsc/v…

Is scientific data ready for its responsible use in AI? Check out our latest paper on @NaturePortfolio Scientific Data. With @JordiCabot and Abel Gómez! nature.com/articles/s4159…

• Social concerns and data limitations are becoming more important. • In the ML community, data is often understood as a “live” asset, of which maintenance policies are a critical part. Check out our preprint: arxiv.org/pdf/2401.10304

Few takeaways: • Profiling those involved in the collection and annotation process is still a weak point inside and outside the ML community • Documenting annotation validation techniques is key to evaluating data quality

Is scientific data 𝗔𝗜-𝗿𝗲𝗮𝗱𝘆? Our latest study in Nature's Scientific Data explores the presence of the ML community's documentation practices in broader scientific data venues and how these venues should adapt their practices to publish AI-ready data. @JordiCabot

@akhtarmubashara @kbontcheva @arkaitz @KingsCollegeLon @kclinformatics @KingsNMES Congratulations!!!

Ending 2024 with exciting news. 🎉 I successfully defended my PhD thesis on Multimodal Automated Fact-Checking earlier this month. Many thanks to my examiners @kbontcheva & @arkaitz, for their thoughtful comments and an enjoyable viva. I am also very grateful for the support I received over the years from my supervisors (@esimperl & Oana Cocarascu), collaborators (@keviv9, @michael_sejr, @vlachos_nlp, @ZhijiangG, @eisenjulian, @chenxipang, @shinde_rajat, et al.), family and friends.

If you are at Neurips, take a look to this. An open format for ML-ready data.

Los españoles🇪🇸 creen que: ➡️Hay más inmigrantes (28%) de los que hay (16%). ➡️Están más desempleados (40%) de lo que realmente están (16%). ➡️Reciben más ayudas (50%) de las que reciben (11%). Contra el relato, datos🤷🏻‍♂️

Avui #23N hem pulveritzat tots els registres: 170.000 persones. La manifestació per l’habitatge més gran de l’història de Catalunya i de l’Estat espanyol! I això només és el principi, anem cap a la vaga de lloguer.

Ep, demà hi ha això: 😉