Modeling consonant-context effects in dialectal variation in a large database of spontaneous speech recordings

Abstract

Given recent interest in the analysis of naturally produced spontaneous speech, we collected, processed, and analyzed a large database of speech samples from the Canadian province of Nova Scotia with the primary aim of examining regional variation in vowel-formant patterns. Although the actual collection of audio recordings is relatively straightforward, the analysis of spontaneous speech suffers from several disadvantages relative to that of laboratory, citation speech: Different vowels have different frequencies of occurrence, surrounding consonants have a large influence on formant peak frequencies, and the distribution of consonant contexts across different vowels is highly unbalanced. To overcome these problems, we developed a statistical procedure inspired by that of Broad and Clermont [1987, J. Acoust. Soc. Am., 81, 155] to estimate the specific effects of both onset and coda consonant-context effects on vowel formant frequencies. However, in contrast to their procedure, both vowel formant frequenc...