L’intelligence artificielle pour préserver le français et l’italien: le projet E-MIMIC

Within the Empowering Multilingual Inclusive Communication project of the Politecnico di Torino, the University of Bologna and the University of Rome 2, we will present tests carried out on the French and Italian versions of the Inclusively device, designed to reformulate administrative texts in an inclusive manner, in order to show how supervised learning can help preserve multilingualism, in particular linguistic diatopic variation. The examples provided allow us to observe how the device resolves problems linked to pre-training on undifferentiated corpora (French case) and the question of the feminization of Romance languages (Italian case). The analysis shows that pre-editing, conceived in a “broad” way, proves fundamental for the preservation of the linguistic diatopic variation.