Saurabh Nath
2025
English-based acoustic models perform well in the forced alignment of two English-based Pacific Creoles
Sam Passmore
|
Lila San Roque
|
Kirsty Gillespie
|
Saurabh Nath
|
Kira Davey
|
Keira Mullan
|
Tim Cawley
|
Jennifer Biggs
|
Rosey Billington
|
Bethwyn Evans
|
Nick Thieberger
|
Danielle Barth
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Expanding the breadth languages used to study sociophonetic variation and change is an important step in the theoretical development of sociophonetics. As data archives grow, forced alignment can accelerate the study of sociophonetic variation in minority languages. This paper examines the application of English and custom-made acoustic models on the alignment of vowels in two Pacific Creoles, Tok Pisin (59 hours) and Bislama (38.5 hours). We find that English models perform acceptably well in both languages, and as well as humans in vowel environments described as ‘Highly Reliable’. Custom models performed better in Bislama than Tok Pisin. We end the paper with recommendations on the use of cross-linguistic acoustic models in the case of English-Based Creoles.
Search
Fix author
Co-authors
- Danielle Barth 1
- Jennifer Biggs 1
- Rosey Billington 1
- Tim Cawley 1
- Kira Davey 1
- show all...
Venues
- acl1