Natural Language Processing (NLP) models have undergone remarkable advances in recent years relying on massive amounts of data compared with humans yet struggling on measures of creativity and generalization. What makes the human processes of language acquisition and learning so powerful? How can what we know about language acquisition and learning inform NLP models, and how can the advances in NLP inform our understanding of the human processes of learning and acquisition? In my talk, I'll present computational analyses of how people acquire language gradually through milestones of errors and breakthroughs. I'll discuss studies suggesting methods to improve language teaching through the fundamental computational processes that enable human learning and proficiency.
Libby Barak is a post-doc researcher in the Computer Science department at Rutgers University - Newark. Her recent research focuses on modeling morphology acquisition by typically developing children and children with developmental language disorders, and predicting the properties of useful language interventions. Before joining Rutgers, Libby was a post-doc researcher in the Psychology department at Princeton University. She received her Ph.D. from the University of Toronto (2016) and M.Sc from Bar Ilan University, both in Computer Science.