PhD Student · University of Turin
I work at the intersection of natural language processing and human reading, combining eye-tracking data with language models. I also contribute to open Transformer models for Italian.
I use eye movements to study reading difficulty and differences between readers. My work includes reader embeddings for personalised reading-time prediction and CERVINO, an Italian eye-tracking dataset of legal texts.
Reader modellingI am a co-first author of AlBERTurin, a fully open family of Italian encoders with modern architectures. My interests include pre-training, model evaluation, and perplexity-based classification.
AlBERTurinEMNLP · 2026
An open resource for studying how readers process Italian legal texts, with eye movements and comprehension measures.
IEEE TAI · 2026
The Delta recalibration strategy addresses bias in paired language-model classifiers, including under imbalanced training data.
CLiC-it · 2026
Co-first author · Equal contribution
Open Italian Transformer encoders, with released models and a shared tokenizer.
Outstanding contribution to Italian NLP resourcesEVALITA · 2026
An Italian BabyLM evaluated on four tasks at EVALITA 2026.
Findings of ACL · 2025
Reader embeddings model individual differences in reading times using eye-tracking data.
2026
CLiC-it 2026 · Best Student Paper Award, Mention of Honor
For AlBERTurin.
2025
IEEE MetroXRAINE 2025
Digital Linguistics · University of Zurich
Department of Computational Linguistics, Switzerland
University of Turin, Italy · CCC group
University of Turin & University of Aosta Valley
Accessibility and readability of legal texts · Oct–Dec 2023 and May–Oct 2024
University of Turin · 109/110
University of Turin · 110/110 with honours
Let's connect
