Eduardo Calò is a PhD candidate in the Natural Language Processing group at Utrecht University. His research focuses on natural language generation (NLG) from semantic inputs, specifically logic-to-text generation, the subfield of NLG that aims to generate optimally intelligible text explaining the meaning of logical formulae. Eduardo is currently part of the NL4XAI framework, a European program aiming to make artificial intelligence more self-explainable and easily accessible to humans. He has a diversified background, holding a BA in Asian studies from La Sapienza and an MSc in natural language processing from Université de Lorraine. Eduardo has also had working experiences in research laboratories (Inria, ATILF) and companies (Samsung) and exchanges in various academic institutions (IIIA, BFSU).
Logical formulae (LFs) are pivotal in many scientific fields, such as Artificial Intelligence and linguistics. Grasping the meaning of LFs is thus crucial for many scholars, yet sometimes even experienced logicians might need help deciphering a complex LF. We can employ natural language generation (NLG) techniques to address this issue. In particular, logic-to-text generation aims to simplify LFs and translate them into clear and understandable natural language text.
In this tech talk, Eduardo introduced this underrepresented area of NLG and illustrate its key challenges. Specifically, he focused on the issues that arise when dealing with such a type of input (e.g., is the shortest LF the best input?) and how to address the evaluation of the generated texts (e.g., what are the appropriate evaluation dimensions to consider?).
He presented the results of his recent research and offer an interactive demo to the audience.