Artificial Intelligence at the service of inclusive language policies: the case of the E- MIMIC Project

Rachele Raus & Tania Cerquitelli

University of Bologna & Politecnico of Torino, Italy

Panel: Language as a means of inclusion in educational and institutional settings

Chair: Maria Margherita Mattioda, Università di Torino, Italy

It is well known how artificial intelligence (AI) learning from big data can contribute to the reiteration of gender bias and forms of exclusion due to the dissemination of stereotyped discourses on minorities, such as migrants and disabled people (Bartoletti 2021, Marzi 2021, Savoldi et alii 2021). The Empowering Multilingual Inclusive comMunICation (E-MIMIC) project led by the Polytechnic of Turin and the University of Bologna, in partnership with the Jean Monnet Centre of excellence Artificial Intelligence for European Integration, aims to promote inclusive communication in real- world scenarios by eliminating non-inclusive language forms in administrative texts written in European countries, starting with those written in Italian and French. The application uses AI algorithms to identify non-inclusive text segments and propose inclusive reformulations. The project starts from the assumption that supervising machine learning through linguistic and discourse criteria can contribute to achieving better quality results. The methodology proposed to identify these criteria rests on the principles of discourse analysis “à la Français” (Dufour, Rosier 2012: 5). In this sense, an attempt is made not to reiterate the non-inclusive ideology present in current discourses (in France and Italy). The application highlights inappropriate segments or words, thus contributing to spreading awareness of discrimination and non-inclusion in language. Moreover, the application suggests possible reformulations, so that the user can choose from the proposed solutions. The AI exploited by the application thus becomes an element in support of linguistic policies that aim at the development of metalinguistic awareness capable of counteracting the circulation of erroneous discursive and linguistic frames, also in the perspective of an eco-critical analysis of discourse (Stibbe 2014). The first tests carried out on the application are encouraging and allow us to extend its implementation to other European languages in addition to Italian and French, taking into account the diatopic variants of the languages analysed.

References

Bartoletti, I. (2021). An Artificial Revolution. On Power, Politics and AI. Edimbourg: Indigo.

Dufour, F., Rosier, L. (2012), Héritages et reconfigurations conceptuelles de l’analyse du discours ‘à la française’ : perte ou profit ?. Langage et Société, 140, 5-13.

Marzi, E. (2021). La traduction automatique neuronale et les biais de genre : le cas des noms de métiers entre l’italien et le français. Synergies Italie, 17, 19-36. http://gerflint.fr/Base/Italie17/marzi.pdf.

Savoldi, B., Gaido, M., Bentivoglio, L., Negri, M., Turchi, M. (2021). Gender Bias in Machine Translation, Transactionsof the Association for Computational Linguistics, 9, 845-874.

Stibbe, A. (2014). An ecolinguistic approach to critical discourse studies. Critical discourse studies, January 2014, DOI: 10.1080/17405904.2013.845789.

Deixa un comentari

L'adreça electrònica no es publicarà. Els camps necessaris estan marcats amb *