Dialect Classification and Speech Intelligibility in Noise
Loading...
Date
2018-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The Ohio State University
Abstract
Listeners have good intuitions about regional differences between speakers. Less studied is the relationship between dialect classification and speech intelligibility. Regional dialect is known to affect speech intelligibility: in some cases, familiar dialects can facilitate speech processing, but in other cases, familiar dialects inhibit speech processing. Many other lexical, discourse, and social factors also affect speech intelligibility. The present study explores predictors of dialect classification accuracy and intelligibility accuracy for the Northern and Midland dialects of American English. In previous studies, Midland speech in noise was found to be more intelligible than Northern speech in noise for both Midland and Northern listeners. Given that listeners are sensitive to differences between Northern and Midland speech in terms of intelligibility, we might expect that listeners use differences in intelligibility to identify where talkers are from. To explore this possibility, participants completed a speech intelligibility in noise task followed by a dialect classification task with Northern and Midland speech. Stimulus materials were balanced for lexical, discourse, and social factors that affect speech intelligibility. A linear regression model revealed that mean token intelligibility did not predict token classification accuracy. The results reveal that easily classifiable tokens are not always the least intelligible. These results suggest that processing mechanisms adapt for regional variation, but dialect-specific cues are not always available to the listener in classification. Predictive models of intelligibility and classification revealed that social, lexical, and discourse factors interact to affect accuracy on both tasks. Ongoing analysis will more precisely determine the role of each predictor in each task, leading towards a better understanding of the relationship between dialect perception and processing mechanisms.
Description
Undergraduate Research Scholarship, $4000, College of Arts and Sciences Honors, for thesis work
Keywords
dialect classification, speech intelligibility in noise, speech processing, dialect