Comparing pre-linguistic normalization models against US English listeners’ vowel perception
DOI:
https://doi.org/10.36505/ExLing-2022/13/0037/000579Keywords:
speech perception, vowel normalization, computational modelAbstract
We investigate the role of pre-linguistic normalization in the perception of US English vowels. Bayesian ideal observer (IO) models were trained on either unnormalized or normalized acoustic cues to vowel identity using a phonetic database of eight /h-VOWEL-d/ words in US English. The predictions generated by the IO models for vowel categorization were then compared with eight-way categorization responses produced by L1 US English listeners in a web-based experiment involving recordings of /h-VOWEL-d/ words. The results indicate that pre-linguistic normalization substantially improves the fit to human responses, increasing performance from 74% to 90% of the best possible fit.
References
Paola Escudero, Bion, R.A.H. 2007. Modeling vowel normalization and sound perception as sequential processes. In Proceedings of the 16th International Congress of Phonetic Sciences (ICPhS 16), 1413-1416.
Richter, C., Feldman, N.H., Salgado, H., Jansen, A. 2017. Evaluating low-level speech features against human perceptual data. In Proceedings of ACL 2017, 5, 425-440.
Xie, X., T. Florian Jaeger 2020. Comparing non-native and native speech: Are L2 productions more variable? The Journal of the Acoustical Society of America, 147, 3322-3347.
Downloads
Published
Issue
Section
License
Articles are published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
