Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Exploring automated formant analysis for comparative variationist study of Heritage Cantonese and English

Cui, Naomi and Zhu, Minyi and Law, Vina and Tse, Holman and Nagy, Naomi (2014) Exploring automated formant analysis for comparative variationist study of Heritage Cantonese and English. In: Change and Variation in Canada / Changement et Variation au Canada (CVC 8), 31 May 2014 - 31 May 2014, Queen's University, Kingston, ON, Canada.

[img]
Preview
PDF
Available under License : See the attached license file.

Download (5MB)
[img] Plain Text (licence)
Available under License : See the attached license file.

Download (1kB)

Abstract

We consider the possibility of Cantonese and English reciprocally influencing vowel space in Toronto’s Heritage Cantonese community by comparing Generation1 and Generation2 speakers in both languages. We predict more English-like patterns in Gen2 Cantonese (vs. Gen1) and more Cantonese-like patterns in Gen1 English (vs. Gen2). Methodological innovations include automated forced alignment and formant extraction for Cantonese -- methods increasingly used for English data but not frequently applied to other languages in sociolinguistics. Extension to additional languages provides testing grounds for sociolinguistic generalizations which have been based primarily on English, French and Spanish. FAVE (Rosenfelder et al. 2011) was used to force-align English transcripts to the corresponding .wav. Cantonese transcripts were force-aligned in ProsodyLab (Gorman et al. 2011), using unsupervised machine learning to train acoustic models, customizable for non-English data (unlike FAVE). FAVE was used to extract and normalize English formant measurements (F1, F2) at each vowel midpoint. A custom Praat script did the same for Cantonese. Data consists of ~40,000 measured vowels for each language: all stressed vowels produced by 10 speakers per language during a 1-hour interview. This paper focuses on ~9,000 tokens of /i/. Preliminary results from mixed-effects modeling: o Generation and sex are main effects in Cantonese for both F1 and F2, but only for F2 in English → As predicted; Gen1 speakers haven’t fully acquired social conditioning in English. Contra predictions, Gen2 sustains Gen1-like social conditioning in Cantonese. o As in Homeland Cantonese (Yue-Hashimoto 1972:158), Heritage Cantonese /i/ shows a centralizing effect of following velars; stronger in Gen1 than Gen2 → supports our hypothesis. Neither generation transfers this effect to English. o Without any human correction, the automatically extracted and measured data behaves much as expected → a promising avenue for further investigation. Comparisons to Toronto Anglo English (Boberg 2008, Roeder & Jarmasz 2010, Roeder 2012) will be reported.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: Conference or Workshop Item (Paper)
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Cui, Naomi
Zhu, Minyi
Law, Vina
Tse, Holmanhbt3@pitt.eduHBT3
Nagy, Naomi
Date: 31 May 2014
Date Type: Publication
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Event Title: Change and Variation in Canada / Changement et Variation au Canada (CVC 8)
Event Dates: 31 May 2014 - 31 May 2014
Event Type: Conference
Institution: University of Pittsburgh
Refereed: No
Official URL: http://cvc8.org/abstracts/
Date Deposited: 26 Oct 2015 15:55
Last Modified: 25 Aug 2017 04:59
URI: http://d-scholarship.pitt.edu/id/eprint/26261

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item