Using forced alignment for sociophonetic research on a minority language

Until recently, large-scale phonetic analyses have been out of reach for under-documented languages, but with the advent of methodologies such as forced alignment, they have now become possible. This paper describes a methodology for applying forced alignment (using the Montreal Forced Aligner) to a speech corpus of Matukar Panau, a minority language spoken in Papua New Guinea. We obtained measurements for 68,785 vowel tokens, produced in both narrative and conversational data by 34 speakers. We examined the social conditioning on a subset of these vowels according to traditional sociolinguistic categories of age and gender, and also consider the impact of clan as a major axis of organization in this community. We show that there is a role for clan as a sociolinguistic factor in conditioning the variation observed.