Speaker: Yang Xu, University of California, Berkeley
Title: Mental Algorithms of the Lexicon and Cognitive Economy
Abstract:
Human language relies on a finite lexicon to express an infinite set of emerging ideas. One result of this tension is that words often acquire novel senses over time. For example, "gay" used to convey a sense of happiness and now also refers to a sexual orientation, and "grasp" was extended from physical to mental acts (as in "grasp an idea"). I present the view that changes like these reflect a drive towards cognitive economy. Previous scholars have suggested that word senses extend in non-arbitrary ways, but no attempt has been made to formalize these proposals and test them against historical data at scale. I provide several such tests, by developing computational models enriched by human judgments that predict the historical development of word senses, dating back hundreds of years. These analyses suggest that word meanings not only extend in predictable ways through mechanisms including semantic chaining and metaphorical mapping, but that these extensions tend to minimize cognitive effort. My approach draws on ideas from psychology, linguistics, philosophy, and natural language processing, and integrates them in an attempt to reverse-engineer the mental algorithms that construct the human lexicon.
Biography:
Yang Xu is a postdoctoral researcher in Linguistics and lecturer of Cognitive Science at the University of California, Berkeley. Being a member of the data science initiative at Berkeley, he has created courses that connect data science with the study of the mind. He obtained his Ph.D. in Machine Learning with a special track in the Center for The Neural Basis of Cognition from Carnegie Mellon University. Previously, he obtained his BA/MEng degrees in Engineering from the University of Cambridge. He is a two-time recipient of the Prize for Computational Modeling of Language from the Cognitive Science Society.
Talk sponsored by: Dept. of Computer Science and School of the Environment
For additional information contact Steve Easterbrook