22 February 2019 - Yang Xu: Seminar
Mental Algorithms of Word Sense Extension
Natural language relies on a finite lexicon to express an infinite set of emerging ideas. One result of this tension is that words acquire novel senses over time. For example, “face” originally referred to a body part and was later extended to convey an expression, the front surface of an object, and defiance of danger (as in “face danger”). Previous scholars have suggested that word senses extend in non-arbitrary ways, but sparse attempt has been made to formalize the cognitive mechanisms that give rise to novel senses and test them against historical data at scale. I present analyses that explore cognitively grounded algorithms of word sense extension and test these in the historical development of word senses, dating back hundreds of years. Our analyses suggest that word senses not only extend in predictable ways, but that they tend do so by reducing cognitive effort mainly through a process of chaining. I discuss the applications and implications of these findings to natural language processing and acquisition.
Yang Xu is Assistant Professor in the Department of Computer Science, jointly appointed with the Cognitive Science Program, at the University of Toronto. His current research concerns computational modelling of language change with a particular focus on characterizing the cognitive mechanisms that underlie the historical emergence of word meanings. He was a postdoctoral researcher in the Department of Linguistics and a lecturer of Cognitive Science at UC Berkeley. He obtained his PhD in Machine Learning from Carnegie Mellon University, and his BA/MEng from the University of Cambridge.