I am a first-year PhD student in Computer Science at the CSE department of the University of Washington. I work on fact verification and its societal challenges. My advisor is Yulia Tsvetkov. My work is currently funded by the DARPA project SemaFor.
Before UW, I was a master student at the Language and Technologies Institute (LTI) at CMU. In 2018, I worked at Microsoft AI Frameworks for a year. I studied physics, math, and computer science at Harvard where I obtained an AB/SM. I am Italian and German, and attended a French school in Rome for fifteen years.
Language and reasoning are intricately related – one is the means through which the other can exist and be expressed. A fascinating aspect about NLP research is that a deeper understanding of language can shed light on human thinking. What ultimately motivates me is the hope of going beyond pattern recognitionand finding linguistically grounded models of language capable of understanding, reasoning, and generating text as effectively and reliably as humans do.
Factuality in NLP
A major problem in natural language understanding is factuality which involves ensuring whether the meaning of a text is consistent in a given context (for example whether the summary is consistent withthe article). I decided to work on factuality as it provides a relatively tractable and well defined setting to approach the overarching problem of reasoning and language understanding. Current NLP methods lack both in ensuring the factual consistency of generated text and detecting factual inconsistencies acrossdocuments. As text generation systems become increasingly more fluent, their practical utility is nowlimited by factual inconsistencies in the generated text, which can make them unreliable and misleading. Furthermore, improvements in the detection of factual errors can have a tremendous social impact, as forexample in helping identify fake news.
Machine learning is rapidly transitioning from research to a core element of modern application development. I hope to bring machine learning tools to a wide audience of developers. While at Microsoft, I worked on developing the machine learning library ML.NET. For more information see our KDD paper.
I love sports. I was on the CMU fencing club. Previously, I was part of the Harvard Fencing varsity team, and competed internationally in modern pentathlon back in Italy.