| Giuseppe's Website

A list of news, mostly to convince myself I haven’t been twiddling my thumbs.

Two papers accepted at EMNLP 2025: stay tuned! 👀
💰 Our proposal to build efficient speech language models was granted funding from the Fundação para a Ciência e a Tecnologia!
💰 Our proposal to study the usability of speech translation systems was granted funding from the European Association for Machine Translation!
Two papers accepted at ACL 2025: Watching the Watchers, our study on gender bias on quality estimation metrics for machine translation, and our study on gender encoding in modern 🎤 speech translation models.
SHADES a large collaborative project collecting and release a large multilingual multicultural benchmark on social steretypes, led by Margaret Mitchell (Hugging Face), was accepted at NAACL 2025.

Two papers accepted at EMNLP 2024 (with awards 🏅): Twists, Humps, and Pebbles, our large study on gender bias of multilingual speech recognition which won the Social Impact Paper Award, and Metrics for What, Metrics for Whom, our position paper on requirements and desiderata of actionability for bias metrics.
🎉 Life update! I moved to Lisbon and started working with André Martins and the Sardine 🐟 group;
Two papers accepted at ACL 2024: Classist Tools, our study on the correlation between social class and performance in NLP, and Building Bridges, a new resource to evaluate gender-fair machine translation to German, which was accepted among Findings.
Our Safety-Tuned LLaMA paper was accepted at ICLR 2024;
Our new approach to interpreting speech classification models was accepted at EACL 2024;

We co-organized the Third Workshop on Safety for Conversational AI at LREC/COLING 2024.
A Tale of Pronouns, our study on mitigating gender bias in MT using interpretability was accepted at EMNLP 2023.
Our project ITALIC, the first large-scale Intent Classification dataset with Italian utterances, got accepted at INTERSPEECH 2023.
Our two research proposals on gender bias of prompted LMs and multimodal humor detection were accepted for Google Research credit funding.
Our study on systematic subgroup divergence of SLU models got accepted at ICASSP 2023. Joint effort with Amazon Alexa AI and Politecnico di Torino.

Moreno and I founded Risorse per la Lingua Italiana (RiTA), an online community to gather Italian NLP researchers and build resources (models, datasets, educational contents) in Italian.
🎉 I successfully defended my Ph.D. at Politecnico di Torino (!).
I co-organized EvalRS, the first data challenge and workshop on rounded evaluation of Recommender Systems, held in conjunction with CIKM 2022. The challenge attracted more than 50 teams from 13 countries from various academic and industry affiliations and was later covered in a Nature Machine Intelligence paper. Joint effort with Stanford University, Microsoft, NVIDIA, South Parks Commons, and Coveo.
I wrote and tutored the NLP tutorials at the Mediterranean Machine Learning School (M2L).
We released ferret an interpretability library that got covered by online tutorials and accepted at EACL 2023 (System Demonstration).
EAR got accepted at ACL 2022 (Findings). With it, one can mitigate lexical overfitting in LLMs/
🎉 I started working as a research associate in the MilaNLP group at Bocconi University, Milan, with Debora Nozza and Dirk Hovy;
Our study on Gender Inclusive Language in Italian got accepted at IEEE Big Data 2022 and later featured at the 8e Congrès Mondial de Linguistique Française. Joint effort with the University of Turin, Bologna, and Bergamo.