Towards a new empiricism for linguistics
Nick Chater, Alexander Clark, John Goldsmith, Amy Perfors · Oxford University Press eBooks · 2015
The central theoretical problem of linguistics is how to develop a general framework which makes it possible to describe grammars for each natural language and justify each grammar with respect to alternative proposed grammars. This chapter explores the way in which an empiricist would approach this problem, dividing the problem into two parts: how well does a given grammar model the data that has been gathered, and how well does the grammar state generalizations that can be discovered both within each language, and just as importantly, across languages? Taking a cue from Minimum Description Length analysis, this chapter argues that the task of developing a general theory of grammar can, and should, be the result of an explicit competition between groups using a common computational architecture, with all agreed on an explicit way of measuring both tightness of fit to data, and of measuring how well generalizations are captured.