gavin abercrombie

📣 Call for Papers: NLPerspectives Workshop at EMNLP 2025! Submission deadline: June 27, 2025. Join us in Suzhou, China (Nov 8th) or online to learn and chat about the latest on approaches in the context of perspectives in NLP research. 📝 #NLPerspectives #EMNLP2025

@_dmh @gkatzia @UilleamUan @AGroundw Meal do naidheachd!

@dirk_hovy Congratulations on metamorphic completion!

@verena_rieser Oops, correct link for the first one: aclanthology.org/2023.semeval-1…

@verena_rieser Can also recommend: SemEval-2023 Task 11: Learning With Disagreements (LeWiDi) aclanthology.org/2023.semeval-1… and Resources for Automated Identification of Online Gender-Based Violence: A Systematic Review aclanthology.org/2023.woah-1.17/ 🙂

Hey, #ACL2023NLP ppl make sure you check out the following papers: Adversarial Robustness of Visual Dialog arxiv.org/abs/2207.02639 The Dangers of trusting Stochastic Parrots arxiv.org/abs/2305.16519 Modelling Disagreement or Modelling Perspectives? arxiv.org/abs/2305.06074

@evanmiltenburg @vinodkpg @Ginger_in_AI @LeonDerczynski @dirk_hovy So far only identical items with a large time interval. Started to look into memory effects but ... it's complicated!

@gavin_does_nlp @vinodkpg @Ginger_in_AI @LeonDerczynski @dirk_hovy Are you analyzing intra annotator agreement only for the same items or also for similar items? E.g. Replacing the subject John with Mary but otherwise keeping the sentence the same, or more complex sets of changes but still fully analogous. That might counter memory effects.

Anyone know of a paper talking about attention and #annotation quality dropping as time increases, or for highly repetitive tasks? (Maybe @Ginger_in_AI or @LeonDerczynski?) #lazyweb #nlproc

@evanmiltenburg @vinodkpg @Ginger_in_AI @LeonDerczynski @dirk_hovy We have some results under review and also some ongoing experiments described here: arxiv.org/abs/2301.10684

@gavin_does_nlp @vinodkpg @Ginger_in_AI @LeonDerczynski @dirk_hovy Also nice! Was it published somewhere?

@vinodkpg @evanmiltenburg @Ginger_in_AI @LeonDerczynski @dirk_hovy Not exactly what we're looking at: we investigate stability/consistency on repeat annotations (2+ weeks later)

@evanmiltenburg @Ginger_in_AI @LeonDerczynski @gavin_does_nlp @dirk_hovy

@fatmaElsafoury @verena_rieser Thanks for coming and giving a great talk!

Today was one of the most exciting days in my PhD. I gave a talk @ the Interaction lab at Harriot-Watt University, met fascinating researchers and had very interesting conversations. Thanks, @verena_rieser and @gavin_does_nlp for inviting me and for the great day.

@prem_k @verena_rieser @aaclmeeting Not a crowd of lay people as the risks are too high. We suggest that qualified emergency medical consultants are needed

@verena_rieser @gavin_does_nlp @aaclmeeting Very interesting implications. However ... Pardon me for channeling James Surowiecki for a moment here ... Would you say that the wisdom of crowds is better than the few experts in this case? Just out of curiosity.

Two new papers on #AISafety and #AIRobustness from my group, which I am really excited about: (1/3)

@mrc_valentino @andrenfreitas @jochenleidner @louiseadennis Congratulazioni!

@federicobianchy @stanfordnlp Wow! Congrats!

Life update! I recently moved to California 🌞 and today is my first day at Stanford! I am joining @stanfordnlp as a postdoc! 🎉🥳

@dotcyril @Ginger_in_AI @verena_rieser @sara_hlt Anywhere on Earth 🌍

@Ginger_in_AI @gavin_does_nlp @verena_rieser @sara_hlt Hi Valerio, a question on the deadline. What timezone is it on?

@_dmh @CreateInf @gkatzia @UilleamUan @InchatD Meal ur naidheachd!

@_dmh 👏👏👏

@vinodkpg @aidaa @blahtino Nice! You may be interested in pdai.info

Thrilled to be able to finally share our work on dealing w/ disagreements in data! Annotators are not interchangeable! When they disagree, especially on subjective tasks, it may be a signal rather than noise. Ongoing work w/ my amazing intern @aidaa & colleague @blahtino. 1/N

@emilymbender @verena_rieser winlp.org/wp-content/upl…

@verena_rieser Thanks, Verena! This one doesn't seem to have a pdf associated -- do you have one somewhere? twitter.com/verena_rieser/…

Great list of papers! Thanks for featuring our work ☺️ Two papers I'd like to add for dialogue/ conversational AI: 1/

@verena_rieser 🤦

Thoughts: Sitting on a panel on "Gender Diversity" at an event where all keynote speakers are white and male? Boycott? Set a positive example? 🤷‍♀️

@jeremy_nlp @upvehu @IxaGroup @ltgoslo Guau! Zelako berri ona, oso oso ondo, zorionakisimo! Egongo gara EHan espero dut. Donostian biziko zara?

@_dmh Something I once did: aclanthology.org/L16-1092.pdf