Elisa Sanchez-Bayona

Come to chat about lemmatization by Gen LLMs today during the Findings session at 12:30! Our joint work with @ragerri and @alan_akbik Paper: aclanthology.org/2025.findings-…

Tomorrow it's poster day at #EMNLP2025! 📝 Meta4XNLI: A Cross-lingual Parallel Corpus for Metaphor Detection and Interpretation @ragerri @Hitz_zentroa 📍 Location: Hall C3 🕓 Time: 10:30-12 Come by to chat about metaphor!

Our weekly seminars are back! 🎉 We welcomed @eli_sanba and @joseba_fdl from the University of the Basque Country. Elisa Sanchez-Bayona discussed Automatic Metaphor Processing, and Joseba Fernandez de Landa Aguirre spoke on instructing LLMs for low-resource languages.

🎉 Happy to present my poster "Metaphor and Large Language Models: When Surface Features Matter More than Deep Understanding" at @aclmeeting in Vienna! 📅 Tue, 29/07 10:30-12:00 Hall 5 (54) Joint work w/ @ragerri @Hitz_zentroa See you there!

3. @eli_sanba , Rodrigo Agerri. Metaphor and Large Language Models: When Surface Features Matter More than Deep Understanding. In ACL Findings, Poster Session 10.30am, 29/07 Tuesday.

In today’s seminar, @eli_sanba presented the work of her ongoing PhD thesis: From Automatic Metaphor Processing in Spanish to a Multilingual Perspective: Annotation, Systems, and Evaluation.

Check out our new accepted paper at @sepln 74: "Language Independent Stance Detection: Social Interaction-based Embeddings and Large Language Models". Joint work with @ragerri at @Hitz_zentroa . arxiv.org/abs/2210.05715

I am happy to announce that our latest work on Visually Augmented Language Models has been accepted in #COLING2025 We present Blind-VaLM, a simple approach removing the need for explicit images to visually augment an LM, while still achieving strong performance on VLU and NLU.

🚨Happy to share that our paper “A LLM-Based Ranking Method for the Evaluation of Automatic Counter-Narrative Generation" has been accepted at #EMNLP2024 Findings! 🎉 Congrats to all coauthors! @Aitor57 @ragerri @Hitz_zentroa @IxaTaldea

today, im excited to release a repository that implements llama3 from scratch -- every matrix multiplication from attention across multiple heads, positional encoding and every other layer in between has been carefully unwrapped & explained. have fun :) github.com/naklecha/llama…

Can you imagine having all the evidence of data contamination gathered in one place? 📢As part of the CONDA workshop, we present the Data Contamination Evidence Collection, a shared task on reporting contamination. Available as a @huggingface space: hf.co/spaces/CONDA-W…

Do LMs struggle to understand the meaning of metaphorical expressions? We tested if LMs correctly perform NLI between a premise and a hypothesis when the inference depends on a metaphor. Results suggest that LMs perform worse when metaphors affect the inference in the NLI task.

Are LMs able to mark which words in a sentence are not literal?  We developed a series of multilingual, cross-lingual and cross-domain experiments to evaluate their capabilities in the task of metaphor identification.

Do LMs correctly interpret metaphor? We present Meta4XNLI! The first EN-ES pallalel dataset annotated for metaphor detection and interpretation via NLI; joint work w/ @ragerri @IxaGroup @Hitz_zentroa 💻github.com/elisanchez-bee… 🤗hf.co/datasets/HiTZ/… 📝arxiv.org/abs/2404.07053

Methapor dataset with some twists They annotated Spanish 117K tokens (3.6K sentences) for being metaphoric or not 🟦new dataset 🟦2-10% of sentences have a metaphor🤨 🟦Cross lingual - SoTA on Spanish, and they didn't even multitask #EMNLP2022livetweet #CoNLL2022 linkes👇

"Leveraging a New Spanish Corpus for Multilingual and Crosslingual Metaphor Detection" #CONLL2022 paper, to be presented by @eli_sanba today 08/12/2022, 11-12.30am, poster board 51 arxiv.org/abs/2210.10358

@Hitz_zentroa @IxaGroup is present in Abu Dhabi to present our work at #EMNLP2022 @eli_sanba @iker_garciaf @aormazabalo @ragerri Please do come to check our work and we will be happy to chat too!

also interesting results on crosslingual transfer of metaphor. Data and code publicly available ixa-ehu.github.io/cometa , accepted as long paper @conll_conf @Hitz_zentroa @IxaGroup paper

Our monolingual and cross-lingual experiments obtain state-of-the-art performance for English and Spanish. Most interestingly, they indicate a high cross-lingual transferability of linguistic metaphor. These results further work is required in this research direction.

CoMeta enables research on metaphor in Spanish but also on cross-lingual approaches to study metaphor and how this form of figurative language behaves in different languages.

We're happy to share our new work "Leveraging a New Spanish Corpus for Multilingual and Crosslingual Metaphor Detection" w/ @ragerri at @Hitz_zentroa that will be presented at #CoNLL2022. Check it out here! arxiv.org/abs/2210.10358