Katerina Margatina
@katemargatina
applying science @aws. final year phd student @sheffieldnlp 🦋 previously spent time @aiatmeta @coastalcph @ecentua #nlproc
ID:3378033143
https://katerinamargatina.github.io/ 15-07-2015 23:36:39
388 Tweets
742 Followers
653 Following
Blog alert on the PRISM alignment project w/ quotes from me & both PhD supervisors (Bertie Vidgen, Scott Hale). As Scott nicely puts it: 'Having more people at the table, having more perspectives represented, ultimately leads to technology that will raise everyone up.'🙌🚀
Happy to share our latest preprint with Andrei Popescu-Belis : 'Don't Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimation'
📝preprint: arxiv.org/abs/2401.06688
Can you pre-train LMs under extreme settings, e.g. only using small subsets of characters from individual tokens instead of full tokens?
Check out our new paper w/ Ahmed Ahmed Alajrami and Katerina Katerina Margatina, to appear at #EMNLP2023
📰: arxiv.org/pdf/2310.17271…
PhD Vacancy Alert!
Join us in spreading the word: Eva Vanmassenhove, afra alishahi and I have a PhD opening on fair and inclusive Machine Translation.
Details & application: tiu.nu/21906
Deadline: Nov 15, 2023
Interviews: Dec 2023
Thanks!
#PhDJob #MachineTranslation
Feels uncomfortable and unimportant to celebrate that our paper got accepted in #EMNLP2023 Findings, given everything that is going on. It was the last paper for my PhD though, so there’s the silver lining!
Rejections are part of research (and life), but this year's EMNLP 2024 review process is very disappointing.
Our paper was rejected despite the meta-review explicitly proposing acceptance.
#emnlp2023
While we're preparing our EMNLP submissions, we're happy to announce that our group has 11 papers (8 main and 3 findings) accepted at #ACL2023NLP Congratulations to all our members and collaborators🥳
🫶 Thrilled to see CAL (arxiv.org/abs/2109.03764 w/ Giorgos Vernikos @loicbarrault Nikos Aletras) as an established AL acquisition method & always happy to see algorithms that build on top of it and improve it! (certainly more to be done still💪)
Very nice work Matthias Aßenmacher!