Measuring phonological distances in a tonal language: an experimental and computational investigation with Cantonese
Youngah Do, Ryan Ka Yau Lai
July 2020

To explore how speakers measure phonological distances in tonal languages, we conducted distance judgement tests of word pairs from Cantonese speakers, then fit Bayesian multilevel models to predict the results from various distance metrics. We find that Hamming distances between segments based on multivalued features and tonal representations incorporating pitch contours, including pitch changes across syllables when applicable, consistently perform best. Moreover, onsets are consistently weighted heavier than codas and tones, although the importance of the nucleus varies between monosyllabic and disyllabic words. We discuss how the current results inform the understanding of phonotactic learning models of tonal languages.
Format: [ pdf ]
Reference: lingbuzz/004369
(please use that when you cite this article)
Published in: To appear in Language
keywords: phonological distances, computational phonology, bayesian modeling, tone, cantonese, phonology
previous versions: v2 [November 2018]
v1 [October 2018]
Downloaded:187 times


[ edit this article | back to article list ]