Pinned Tweet
Tales from the ISLab is now online! Check out the latest articles on our new blog: tales.islab.di.unimi.it
Once upon a time… #talesfromtheISLab
Un invito a ponderare rischi e opportunità e trarre il meglio dalla rivoluzione #IA in corso, portando un reportage della #MilanoDigitalWeek legandolo a ciò che facciamo quotidianamente. Un grazie speciale agli organizzatori di tutti gli eventi citati!
tales.islab.di.unimi.it/2024…
islab-unimi retweeted
Our survey paper on modeling #SemanticChange using #LLMs has officially been accepted by #ACM Computing Surveys! It was a challenging journey! Huge thanks to my @IslabUnimi co-author/supervisor, and everyone who supported me. Happy reading!
doi.org/10.1145/3672393
#NLP
islab-unimi retweeted
Our latest preprint is now available at arxiv.org/pdf/2404.18570. We propose a replacement schema to simulate different semantic changes and use it to analyze semantic change through the lens of 𝐋𝐚𝐫𝐠𝐞 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬 such as BERT, mBERT, XLM-R, and LLaMa 2. #NLP
🎉 We are excited to share that our latest research has been accepted at LT4HALA @LrecColing!
📃 Check out the pre-print here: buff.ly/4aIdVTH
👋 See you there on May 25th!
islab-unimi retweeted
Excited to share that our paper has been accepted to the main NAACL conference! 🎉 Check out the preprint here: arxiv.org/abs/2402.12011
#NAACL2024 #NLP #LSC #SemanticChange
Have you ever wondered what tools like ChatGPT have changed in the realm of AI research? Spoiler: it has been a massive shock for many! We talk about it in our first blog post in Italian.
buff.ly/3PbJbSa
islab-unimi retweeted
🎬 It's a wrap! @projectVAST partners at final project review in Brussels, which concluded with great success! For the past 3 years, we worked meticulously on #values #digitisation #culturalheritage and development of tools. Expressing gratitude to partners for their hard work!
islab-unimi retweeted
Excited to announce our latest research: A systematic comparison of contextualized word embeddings for modeling Lexical Semantic Change.
Check out our paper: arxiv.org/pdf/2402.12011.pdf
#WordinContext #WordSenseInduction #LexicalSemanticChange #ContextualizedEmbeddings #GPT4 #LSC
islab-unimi retweeted
.@projectVAST runs an experimental activity, involving 8 primary school classes from “Scuola Primaria Morosini” in Milan, involving 160+ children focusing on “Little Red Riding Hood” #fairytale designed by teachers with support from @LaStatale VAST team!
vast-project.eu/the-vast-pro…
Hi all! We have just published a new PyPI package called ✨relearn✨.
It's a starter kit for Reinforcement Learning projects, and we have been using it for studying algorithms from Sutton and Barto.
- GitHub: buff.ly/3wigwUM
- PyPI: buff.ly/3SxfZpB
By marking sample phrases and vocabulary as modern, historical or neutral, we were able to compare what historical tendencies models represent. We compared #MacBERTh, BERT-based English #HistoricLanguageModel and BERT base.
We confirmed that MacBERTh was best able to represent Early Modern English texts. Unsurprising, as MacBERTh it is the only one that includes #EEBO, the major corpus representing the era, in its training set. Yet, this serves as a validation of the method of comparison.
Happy to share a pre-print examining how well #BERT-based models can represent #EarlyModernEnglish text. We used a previously-used bias-detection approach to compare the ability of models to predict language variants from the past
buff.ly/49qa8t8
@sednolimodo @mapto
Within the @projectVAST, we have developed and interactive book. In it #LittleRedCap, we engage children with questions to both encourage them to think critically of the classical tale, but also to get an aggregated view of how it is perceived today
buff.ly/3HGPd9q
islab-unimi retweeted
We recently proposed an 𝘪𝘯𝘤𝘳𝘦𝘮𝘦𝘯𝘵𝘢𝘭 extension of Affinity Propagation based on Cluster Consolidation and Stratification. While waiting for an official publication, check out our arXiv preprint: arxiv.org/abs/2401.14439 #clustering
islab-unimi retweeted
Is ChatGPT the ultimate solution for modeling Semantic Change? Check this: arxiv.org/pdf/2401.14040.pdf
We show that: when used as an off-the-shelf model, with a reasonable prompt from a human perspective, the current (Chat)GPT doesn't outperform pre-trained BERT #ChatGPT #BERT #LSC
islab-unimi retweeted
I am very happy to announce that we got a paper accepted at Findings of EACL 2024 in Malta! Stay tuned for more details. Looking forward to seeing you there! 2024.eacl.org
Don't forget to attend our tutorial on Modeling lexical semantic change! changeiskey.org/event/2024-e…
islab-unimi retweeted
Excited to share insights on Computational Modeling of Semantic Change at the LOT Winter School in Tilburg! Check out our yesterday's session 👇
docs.google.com/presentation…
Today, more insights from the CIK team. Tomorrow, hands-on sections!
changeiskey.org/
islab-unimi retweeted
Today I held a talk at the University of Gothenburg about my research with the #ChangeIsKey and #ISLab team. Check out my slides and references to learn more!
Title:
Studying Word Meaning Evolution: Incremental Lexical Semantic Detection
Slides: docs.google.com/presentation…
islab-unimi retweeted
EACL 2024 (1/3) @eaclmeeting
- Computational modeling of semantic change. Pierluigi Cassotti, Francesco Periti, Stefano De Pascale, Haim Dubossarsky and Nina Tahmasebi.
- Language + Molecules. Carl Edwards, Qingyun Wang and Heng Ji. #NLProc










