Linguaggio inclusivo e intelligenza artificiale: il progetto E-MIMIC e la disabilità

The developments regarding linguistic inclusivity and the guidelines produced so far demonstrate the growing need and commitment to overcome the use of discriminatory terms and expressions within discourse. This thesis aspires to contribute to the debate on linguistic inclusivity and offer new solutions to those produced so far by providing inclusive rewriting options not only centered on gender but also on other sources of discrimination, without excluding an intersectional perspective. Specifically, it examines the use of non-inclusive language within Italian administrative documents, focusing on the representation of people with disabilities and utilizing the current artificial intelligence tools to promote more inclusive language choices. In this context, the thesis aims to propose the implementation of an application for the inclusive reformulation of administrative texts. This application, which was developed within the E-MIMIC project by the Polytechnic University of Turin and the University of Bologna, has so far been trained exclusively for the recognition of gender biases in order to reformulate them inclusively. Therefore, we considered incorporating into the application the part related to the discrimination concerning disability. The results of the tests conducted on E-MIMIC using the text annotations and disability data produced in this thesis provided excellent insights for future interventions aimed at refining the classifying and reformulating abilities of the application.