Diego Sáez Trumper

🏢 #CIKM2025 Call for Industry Day Talks! Share how your organization applies ML, IR, and data tech in real-world settings. We welcome startups to enterprises – focusing on technical challenges, not product pitches. 📅 Submission: June 27 🔗 Details: cikm2025.org/calls/industry…

📢Don't miss the deadlines! More information: cikm2025.org

BERT is so back 🔥 @answerdotai and @LightOnIO released ModernBERT lightning-fast state-of-the-art BERT model with Apache 2.0 license 🥹 twice as fast as debertav3 on short text and thrice faster than nomic on long text 💨

@FerrarelliM @e__migrante 🗨️🌐 En el marco de una nueva edición de #Wikipuentes, Wikimedia Argentina junto a la Universidad Nacional de La Plata conversamos con @FerrareliM y Diego Saez Trumper sobre IA, cultura digital y Wikipedia. >> youtube.com/watch?v=CNHpju…

👩🏽‍🏫En el marco de una nueva edición de #Wikipuentes, Wikimedia Argentina junto a la Universidad Nacional de La Plata, invitan a un conversatorio abierto sobre Inteligencia Artificial en Wikipedia y la cultura digital. >> wikimedia.org.ar/2023/11/07/ia-…

Attending @trokhymovych's Phd proposal defense. Proud of co-advising his work and sharing this with @PolarBearby. You can check some outcomes: * "Fair multilingual vandalism detection system for Wikipedia" (KDD'23) arxiv.org/pdf/2306.01650 * And more: scholar.google.com/citations?user…

Diego Saez-Trumper (@e__migrante) & Pablo Aragón (@elaragon) from the @Wikimedia Research team will present on their work: "Multilingual approaches to support knowledge integrity in Wikipedia" paper: arxiv.org/pdf/2306.01650…

"Fair multilingual vandalism detection system for @Wikipedia" (Trokhymovych et al, 2023) @PolarBearby

Our paper "Fair multilingual vandalism detection system for Wikipedia" (KDD'23) is now available! Here we describe our last model to fight #vandalism in #Wikipedia! Paper: arxiv.org/pdf/2306.01650… Code: github.com/trokhymovych/K…

"Longitudinal Assessment of Reference Quality on Wikipedia", measuring English Wikipedia reference quality via: - Reference Need, the % of sentences missing a citation. - Reference Risk, the % of non-authoritative references. ( Baigutanova et al, 2023) arxiv.org/pdf/2303.05227…

Join us for the next ISC / @Wikimedia #webinar on safeguarding sources of scientific information online, where we will explore the #Wikipedia model and learn about the @HacksHackers organization. 📅 16 March ⏱️ 14:00 CET | 13:00 UTC Register at: council.science/events/managin…

@skfelipe1 @charles_irl Which kind of embeddings? There are several MLM trained on Wikipedia, but I guess you are referring to something more specific, true?

@charles_irl @e__migrante

Wikipedia publishes official XML dumps, which make it easy to build a database over Wikipedia content -- ran into this in a "hello world" project for Apache HBase. Would be great to see "official" embeddings too, to improve and standardize LM access & vectorDB construction.

@GabrielMagno @virgilioalmeida @TheWebConf Congrats @GabrielMagno and @virgilioalmeida ! great work! Hopefully see you at WWW this year ;)

"Crosslingual Section Title Alignment in @Wikipedia", a probabilistic method that identifies cross-language mappings based on section content features. (@dedwit et al, 2022) dlab.epfl.ch/people/west/pu… @e__migrante @cervisiarius

It was an honor to be invited to present our work at Responsible AI track at the Korean AI summit in Seoul. Thanks @nekozzang for inviting me! Looking forward to keep collaborating with you and your amazing students! Find the full presentation here: youtu.be/oQfNZECt-uc?t=…

📢Ante las voces que cuestionan el rol de #Wikipedia y la calidad de su información, desde @wikimedia_cl queremos aprovechar de recordar cómo funciona la enciclopedia más leída del mundo y la comunidad que la hace posible. 🧵Abrimos hilo: fayerwayer.com/internet/2022/…

Les companyes de Balkumbia acaben de traure disc !! Vos el deixem ací per a que viatgeu amb les músiques tan boniques que fan :) youtube.com/watch?v=7f3xY_… #donaliAlPlay #Balkúmbia #seguiuAestaGentQueFanBonaMandanga

Las #40Horas son una buena noticia!! @e__migrante nos comparte lo que está pasando en este minuto con el anuncio del presidente @gabrielboric. Avanzar en el proyecto es sin duda una mejora en la calidad de vida para las familias, las y los trabajadores.

@gboccardobosoni @gabrielboric Dale subse!

@espejolento @wikidata Gracias por la difusión :) Sería genial presentar estos mismos resultados en Castellano!