Yulia Tsvetkov

My research targets practical problems in Natural Language Processing — the subfield of Artificial Intelligence focusing on computational processing of human languages. I am particularly interested in hybrid solutions at the intersection of machine learning and theoretical or social linguistics, i.e., solutions that combine sophisticated learning methods and insights about human languages or about people speaking these languages. Much of my work focuses on multilinguality, with a unifying goal to extend capabilities of human language technologies beyond individual language boundaries, and to enable NLP for resource-constrained languages, the languages that need it most. 

My problems of interest currently include, but are not limited to: 

• Modeling linguistic phenomena — phonology, morphology, semantics, and pragmatics — across languages, for cross-lingual transfer learning
• Deep learning models for resource-constrained languages
• Interpretable deep learning models for computational social sciences 
• Learning and evaluation of distributed representations of words and phrases 
• Learning to optimize curricula of non-convex models
• Machine translation of text and speech