Fixing Rogue Memorization in Many-to-One Multilingual Translators of Extremely-Low-Resource Languages by Rephrasing Training Samples
- Paulo Rodrigo Cavalin
- Pedro Domingues
- et al.
- 2024
- NAACL 2024
Paulo Cavalin is a Research Scientist of the Conversational Intelligence Group, at IBM Research - Brazil, conducting both theoretical and applied research in Machine Learning, with particular focus on Natural Language Processing problems such as text classification and machine translation for conversational systems. Currently he is working with Foundation Models, focusing on understand their applicability for endangered, very-low-resource, languages such as Brazilian Indigenous Languages.
He holds a Ph.D. degree in Automated Production Engineering/Computer Science from École de Technologie Supérieure (ETS) - Université du Québec, Montreal (QC) - Canada, obtained in 2011, and 15+ years of experience in research in AI-related areas such as Machine Learning, Pattern Recognition, Computer Vision, and Social Data Analytics.
He is also an author of dozens of peer-reviewed scientific papers and an inventor of several patents. A detailed list of can be found at Google Scholar profile.