Perspective-Aware NLP
2024–2026
This University of Turin and Amazon Alexa research collaboration examines how individual and group perspectives affect the interpretation and generation of pragmatically constrained language.
My role involved developing multilingual corpora, annotation guidelines and end-to-end LLM evaluation pipelines for irony, rhetorical figures and generational language variation. I compared prompting and fine-tuning strategies under aggregate and perspective-specific settings, combining automatic metrics with human-centred evaluation.
NoVAGraphS: Non-Visual Access to Graphical Structures
2021–2024
NoVAGraphS investigated how conversational technologies can make graphical and formal STEM content accessible to blind and visually impaired learners.
My role involved designing and evaluating accessible dialogue systems for the non-visual exploration of mathematical expressions, functional diagrams and finite-state automata. I contributed to the dialogue architecture, language-understanding components, interaction data and user-centred evaluations that translated accessibility needs into successive system improvements.
Project details
Code and research resources
My public repositories collect selected code, datasets and reproducible experimental resources developed across research on dialogue systems, natural language generation, accessibility and perspective-aware NLP.
Explore GitHub