This volume documents the Proceedings of the 2nd Workshop on Perspectivist Approaches to Disagreement in NLP, held on September 30th as part of the ECAI 2023 conference (26th European Conference on Artificial Intelligence ECAI 2023) in Kraków, Poland. Until recently, the dominant paradigm in natural language processing (and other areas of artificial intelligence) has been to resolve observed label disagreement into a single “ground truth” or “gold standard” via aggregation, adjudication, or statistical means. However, in recent years, the field has increasingly focused on subjective tasks, such as abuse detection or quality estimation, in which multiple points of view may be equally valid, and a unique 'ground truth' label may not exist. At the same time, as concerns have been raised about bias and fairness in AI, it has become increasingly apparent that an approach which assumes a single “ground truth” can erase minority voices. Strong perspectivism in NLP [1] pursues the spirit of recent initiatives such as Data Statements[2], extending their scope to the full NLP pipeline, including the aspects related to modelling, evaluation and explanation. The Workshop on Perspectivist Approaches to NLP explores current and ongoing work on the collection and labelling of non-aggregated datasets, and approaches to modelling and including these perspectives, as well as evaluation and applications of multi-perspective Machine Learning models. The first edition was held at the Language Resources and Evaluation Conference (LREC) in Marseille in 2022. In this second edition, the workshop received 11 submissions, including nine research papers (one of which non-archival) and two research communications. Of these, nine contributions were accepted. The proceedings are composed of the six accepted archival research papers.
• Agostina Calabrese, University of Edinburgh (Scotland) • Amanda Cercas Curry, Bocconi University (Italy) • Teddy Ferdinan, Wrocław University of Science and Technology (Poland) • Annette Hautli-Janisz, University of Passau (Germany) • Ali Hürriyetoğlu, Royal Netherlands Academy of Arts (Netherlands) • Cassandra L. Jacobs, University at Bufalo (United States) • Anna Koufakou, Florida Gulf Coast University (United States) • Sofie Labat, Ghent University (Belgium) • Marta Marchiori Manerba, University of Pisa (Italy) • Michele Mastromattei, University of Rome “Tor Vergata” (Italy) • Massimo Poesio, Queen Mary University of London (United Kingdom) • Manuela Sanguinetti, University of Cagliari (Italy) • Zeerak Talat • Tiago Timponi Torrent, Federal University Juiz de Fora (Brazil) • Alexandra Uma, Builder.ai (United Kingdom) • Tharindu Cyril Weerasooriya, Rochester Institute of Technology (United States) This workshop was funded through a donation from Amazon. Furthermore, NLPerspectives is organized partially in the framework of the following research projects: • EPSRC ‘Equally Safe Online’ (EP/W025493/1) • EPSRC Designing Conversational Assistants to Reduce Gender Bias (EP/T023767/1) • Kick-of preventIng and responDing to children and AdolesCenT cyberbullyIng through innovative mOnitoring and educatioNal technologieS (KID_ACTIONS) • Compagnia di San Paolo -Bando ex-post 2020 – “Toxic Language Understanding in Online
Communication – BREAKhateDOWN”