Word meanings and human cognition: A computational perspective
Speaker: Yang Xu, University of California at Berkeley
Location: Warren Weaver Hall 1302
Date: April 6, 2016, 11:30 a.m.
Host: Dennis Shasha
Every language has its own unique way of specifying meaning. I explore three problems that concern the relationship between word meanings and human cognition from a computational perspective. Firstly, word meanings vary across languages --- how is this variation cognitively constrained? Secondly, word meanings vary over time --- what principles support the evolution of word meanings? Finally, word meanings interface with cognition --- how do they affect cognitive functions such as memory? I address these problems in three semantic domains: spatial relations, household containers, and color. By applying computational models to rich data sources, I show that 1) despite cross-language variation, word meanings may be constructed from a set of universal primitives; 2) word meanings tend to evolve from a process of chaining while conforming to the principle of efficient communication; 3) the influence of word meanings on cognition can be cast as probabilistic inference. I will finish by discussing future challenges and how an interdisciplinary effort that draws on machine learning, data science and neuroscience can contribute to a computational account of semantics and cognition.
Yang Xu is a postdoctoral researcher in the Department of Linguistics and lecturer of the Cognitive Science Program at the University of California, Berkeley. His primary research is in the area of language and cognition where he develops computational models to study the relations between word meanings and human cognition. His work draws on methods from machine learning, probabilistic modeling, information theory, and computational linguistics to test theories against data from cross-language resources, behavioral experiments, historical corpora and lexicons. He has received twice the Prize for Computational Modeling in Language from the Cognitive Science Society. His most recent work focuses on predicting how word meanings might have evolved over history.
Yang obtained his Ph.D. in Machine Learning from Carnegie Mellon University in 2013. There he developed statistical machine learning methods for MEG brain imaging and studied the neural basis of human visual category learning. Previously, he had obtained his master's and bachelor's degrees in Engineering from the University of Cambridge in the U.K.
At Berkeley, Yang develops and teaches the undergraduate course on "Data Science and The Mind" that engages computational and statistical data analyses with theoretical ideas in cognitive science. He endorses diversity and is a trilingual speaker of English, Mandarin Chinese and Shanghainese.
In-person attendance only available to those with active NYU ID cards.