Spatial Modeling of Object Data: Analysing Dialect Sound Variations Across the UK

• Activity Number: 2 - Phase Variation (Curve Registration) in Functional Data Analyzing
• Type: Invited
• Date/Time: Sunday, July 30, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Shahin Tavakoli* and Davide Pigoli and John AD Aston
• Companies: University of Cambridge and University of Cambridge and University of Cambridge
• Keywords: acoustic linguistic data ; object data analysis ; phase misalignment ; non-parametric smoothing

Abstract

Dialect variation is of considerable interest in linguistics and other social sciences. However, traditionally it has been studied using proxies (transcriptions). When analysing acoustic recordings directly, the problem of phase misalignment arises, since speakers can speak faster or slower, and this is particularly true when data are collected from different sources where the context is different, as it is usually the cases for the large phonetic corpora that are needed to study dialect variation. Moreover, it is also possible for dialect variation to influence speech rate, longer or shorter vowels being often discriminative dialect features. We show how these issues relate to the reconstruction of smooth spatial variation for the sounds data, focusing in particular on the spoken part of the British National Corpus, deposited at the British Library, London. We produce maps of the dialect variation over Great Britain and reconstruct the pronunciation change across the domain of interest.