Modeling phonetic variation in pluricentric languages: an integrative approach

Adaptive speech technologies offer a vehicle for representing pluricentric language variation and the description of both dominant and non-dominant speech varieties. In this article, the work of the LUPo project is described for modeling phonetic variation across national and sub-national varieties of the Portuguese language. While the motivation for this research is based around the development of highquality pronunciation lexica for a Portuguese text-to-speech system – a goal which, itself, is aimed at facilitating the entry of lesser or undocumented variants into the digital domain – the repercussions for pluricentricity are far reaching. We describe how systems such as LUPo can be used to model variation across phonetically similar and disparate national, sub-national, and sociolectal varieties, as well as presenting linguists with a means of testing and observing notions of linguistic distance in terms of shared or innovative rules and phonetic features, and for evaluating the pulling effect of different linguistic centers.