Abstract

Artificial intelligence is opening up new possibilities in education and speech and language therapy, especially for learners with special educational needs and disabilities. Many of these learners improve in word recognition, but reading comprehension often remains slow, effortful, and frustrating—affecting not only school performance but also everyday life. In this talk, I will present an approach that combines psycholinguistic theory with natural language processing to better support reading comprehension. The idea is to take what we know about how people understand language, model these processes computationally, and turn them into practical tools for education and rehabilitation. I will introduce a set of NLP methods designed to target different aspects of comprehension, including sentence simplification, coherence classification, keyword extraction, text-to-pictogram translation, and concept map generation. Together, these methods form a modular toolbox, with particularly promising results in simplifying texts and generating concept maps to support understanding. I will then show how these components come together in an interactive online platform, developed with input from educators and clinicians and evaluated through user studies. These studies highlight the potential of AI-driven tools to provide meaningful support for neurodiverse learners. Overall, the talk offers a perspective on how AI can move beyond isolated tools toward integrated, theory-driven solutions.

Short Bio

Martina Galletti is currently a researcher in artificial intelligence (NLP) at Sony Computer Science Laboratories in Paris, where she works on the development and evaluation of systems based on language models. She recently completed a PhD in artificial intelligence at Sapienza Università di Roma, focusing on user-centered AI, human–machine interaction, and NLP applications in healthcare and education. Her work lies at the intersection of NLP, LLMs, and the humanities, with a particular focus on the accessibility and evaluation of AI systems in real-world contexts.