Abstract

The Harvard Dialect Survey of 2002-3 represented the first linguistic foray into large-scale crowdsourcing (60K respondents) incentivized by dynamic geospatial imaging. Working in tandem with statistics graduate student Josh Katz of North Carolina State University I expanded this in 2013 to make the New York Times dialect quiz, which deployed Josh’s brilliant tweaks of existing clustering, visualization, and prediction algorithms to attract responses to my survey questions from more than 21 million humans. Since that time I have been collaborating with forensic linguist Jack Grieve of Aston University to extract linguistically-significant patterns and trends from our megacorpus. In this talk I report on the development of the New York Times quiz and some of the leading discoveries that have emerged from it, including isogloss conspiracies and stability, the role of political and commuting zones, and multivariate non-local cultural regions.