Xinyu Zhang


Xinyu Zhang's Page

Bio

I am Xinyu Zhang*, a linguist in training. I currently study at CLS Radboud and International Max Planck Research School for language sciences.

Previously I studied at ACLC of University of Amsterdam, under the supervision of Dr. Rob van Son and Dr. Paul Boersma. My main areas of interest are phonetics and phonology, acoustic and articulatory phonetics, and production-perception interaction.

My most recent research projects include a study on the cross-generation auditory dispersion of the Dutch sibilants /s/ and /ɕ/, and a study of the psychoacoustics of perceived voice qualities after the treatment of laryngeal carcinoma.

Please like and subscribe :)

* Prescribed pronunciation: [ɕin˥˥ɥy˨˩˦ ʈ͡ʂɑŋ˥˥], but I welcome and am interested in all phonological adaptations.

Education

Radboud University /

IMPRS Graduate School Max Planck Institute for Psycholinguistics | Nijmegen, The Netherlands

PhD in Linguistics | September 2021 - now

University of Amsterdam - ACLC | Amsterdam, The Netherlands

rMA in Linguistics | September 2019 - 2021

University of Amsterdam | Amsterdam, The Netherlands

MA in General Linguisitics | September 2018 - August 2020

Party Tricks

Praat sings Happy Birthday | Praat Script | © Xinyu Zhang, 2019

Let R automate participant notification email for ya | R script + Rmd|© Mine Çetinkaya-Rundel, 2015; modified by Xinyu Zhang, 2021

Results from the voice quality study:

Progress as of March 3, 2021

  • View slides

  • Some highlights:

Roughness as predicted by acoustic measurements

Roughness as predicted by acoustic measurements

De-correlated breathiness as predicted by acoustic measurements

De-correlated breathiness as predicted by acoustic measurements

Results from the sibilant study

(All participants who indicated interest in the findings on the consent form have been emailed the full results. Thank you all for donating your voices <3)

If you want to read the whole paper: PDF.

Otherwise, some TL;DR:

Acoustic distance between [ɕ] and [s] in each participant, separated by age group:

Speakers aged between 18-30, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

Speakers aged between 18-30, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

Speakers aged above 60, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

Speakers aged above 60, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

The red and blue circles in the above plot for older speakers have more overlap than in the first plot for young speakers.

Alternatively, if we run a PCA on energies by frequency bins:

“Elements” in the following plots refer to frequency bins in the long-term-average-spectra (bin width = 250 Hz, LTAS range = 550 - 10000 Hz).
Eigenvector 1 looks like it's just amplitude difference

Eigenvector 1 looks like it’s just amplitude difference

If we then plot eigenvector 2 against 3 by age group:
age > 60

age > 60

18< age < 30

18< age < 30

** Or, We can also try some random forest!**
ranking of bins in the younger speakers

ranking of bins in the younger speakers

ranking of bins in the older speakers

ranking of bins in the older speakers

Things you didn’t need to know:

If you want to get to know me a tiny bit better, here are my cardinal vowels plotted.

(I mean, what better way to get to know someone, right?)

Also, I play the sacriligious otamatone

Xinyu Zhang


Xinyu Zhang's Page

Bio

Xinyu Zhang ([ɕin˥˥ɥy˨˩˦ ʈ͡ʂɑŋ˥˥]) is a linguist in training. She currently studies at CLS Radboud and International Max Planck Research School for language sciences.

Previously she studied at ACLC of University of Amsterdam, under the supervision of Dr. Rob van Son and Dr. Paul Boersma. Her main areas of interest are phonetics and phonology, acoustic and articulatory phonetics, and production-perception interaction.

Her most recent research projects include a study on the cross-generation auditory dispersion of the Dutch sibilants /s/ and /ɕ/, and a study of the psychoacoustics of perceived voice qualities after the treatment of laryngeal carcinoma.

Please like and subscribe :)

Education

International Max Planck Research School for language sciences | Nijmegen, The Netherlands

Graduate student | September 2021 - now

Radboud University | Nijmegen, The Netherlands

PhD in Linguistics | September 2021 - now

University of Amsterdam - ACLC | Amsterdam, The Netherlands

M.Phil in Linguistics | September 2019 - 2021

University of Amsterdam | Amsterdam, The Netherlands

MA in General Linguisitics | September 2018 - August 2020

Party Tricks

Praat sings Happy Birthday | Praat Script | © Xinyu Zhang, 2019

Let R automate participant notification email for ya | R script + Rmd|© Mine Çetinkaya-Rundel, 2015; modified by Xinyu Zhang, 2021

Results from the voice quality study:

Progress as of March 3, 2021

  • View slides

  • Some highlights:

Roughness as predicted by acoustic measurements

Roughness as predicted by acoustic measurements

De-correlated breathiness as predicted by acoustic measurements

De-correlated breathiness as predicted by acoustic measurements

Results from the sibilant study

(All participants who indicated interest in the findings on the consent form have been emailed the full results. Thank you all for donating your voices <3)

If you want to read the whole paper: PDF.

Otherwise, some TL;DR:

Acoustic distance between [ɕ] and [s] in each participant, separated by age group:

Speakers aged between 18-30, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

Speakers aged between 18-30, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

Speakers aged above 60, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

Speakers aged above 60, Center of Gravity on X-axis, standard deviation of the peak on Y-axis

The red and blue circles in the above plot for older speakers have more overlap than in the first plot for young speakers.

Alternatively, if we run a PCA on energies by frequency bins:

“Elements” in the following plots refer to frequency bins in the long-term-average-spectra (bin width = 250 Hz, LTAS range = 550 - 10000 Hz).
Eigenvector 1 looks like it's just amplitude difference

Eigenvector 1 looks like it’s just amplitude difference

If we then plot eigenvector 2 against 3 by age group:
age > 60

age > 60

18< age < 30

18< age < 30

** Or, We can also try some random forest!**
ranking of bins in the younger speakers

ranking of bins in the younger speakers

ranking of bins in the older speakers

ranking of bins in the older speakers

Things you didn’t need to know:

If you want to get to know me a tiny bit better, here are my cardinal vowels plotted.

(I mean, what better way to get to know someone, right?)

Also, I play the sacriligious otamatone