Automatic Detection of Voice Onset Time Contrasts For Use in Pronunciation Assessment

1
Automatic Detection of Voice Onset Time
Contrasts For Use in Pronunciation Assessment
Abe Kazemzadeh1, Joseph Tepperman1, Jorge Silva1,
Hong You2, Sungbok Lee1, Abeer Alwan2, and
Shrikanth Narayanan1 University of Southern
California1 and University of California Los
Angeles2
Project Description

• Automatically distinguish whether a voiceless
  stop consonant is pronounced with a native or
  accented pronunciation based on voice onset time
  (VOT) characteristics.
• Use data from the Tball corpus ESL children
  doing oral reading tasks.
• Evaluate different methods of accomplishing this.
• State duration measurements
• Explicit modelling of aspiration
• Phone probability discrimination

Results
Methodology
Motivation for Studying VOT

• This study was motivated by a desire to determine
  if a phone was pronounced with a non-standard
  pronunciation.
• Other reasons to study VOT
• It is an important contrastive feature
• It gives information about stress
• It gives information about word segmentation
• It may give information about emphasis

• Baseline method error rates
• p 55 t23 k29
• p 19 t20 k48 using duration of 3rd HMM
  state
• With aspiration model
• ShortVOT/ LongVOT
• p 5 / 36
• t 11 / 38
• k 57 / 17
• With probability comparison
• p 36 / 4
• t 0 / 5
• k 0 / 6
• (trained on test dataover trained?)

• Baseline use duration measurements from a forced
  alignment.
• Insert an /h/ symbol in the transcriptions with
  standard pronunciation, train accordingly and
  decode the test files to see if the /h/ phone is
  recognized.
• Cut out the phones of interest from the audio
  file, train separate models and a combined model,
  and evaluate the likelihood of the separate
  models with respect to. the combined model.
• The data was transcribed by ear with special
  symbols for non-standard pronunciations
• Standard 3 state HMMs.
• The evaluation metric used was the error rate for
  both classes evaluated separately.
• When using thresholds, the point of equal error
  rate for both classes was used.

What is VOT?

• It is the interval between the release of
  closure of an articulator (the transient
  burst) and the start of voicing.
• Defined for stop consonants, e.g. /p,b,t,d,k,g/
• VOT has a continuum of values
• When the start of voicing precedes the release of
  closure for a stop, VOT is negative.
• When the release of closure and onset of voicing
  are coincident, VOT is zero.
• When voicing comes after release of closure, VOT
  is positive.

Tball Corpus

• Los Angeles area elementary schools.
• 256 Children, mainly Spanish native speakers.
• Reading words, letters, and numbers, and naming
  pictures and colors.
• Collected by cooperation between USC and UCLA.

Discussion
Physical Realization of VOT

• Studies have noted that for VOT kgttgtp.
• Roughly, each method increased in difficulty.
• The results improved from the baseline, but the
  last approach (comparing probabilities) may have
  been over-trained.
• Comparing probabilities may be easier to extend
  to other pronunciation modelling tasks.
• Increasing the frame rate didn't help much.
• If an Initial consonant has a short VOT, this
  does not necessarily imply non-standard accent,
  one must know the stress pattern of the word.

• Stop consonants are produced with a closure of
  the vocal tract at a specific point, the place of
  articulation
• During the closure, there is a build up of
  sub-laryngeal pressure.
• When the closure is released there is a transient
  burst of air, frication due to turbulence at the
  place of articulation, aspiration noise from
  turbulence at the glottis
• Voicing may occur before, during, or after the
  release of closure.

tester
techie
child
Conclusion
Linguistic Significance of VOT

• When classifying stop consonants based on VOT
  characteristics, different approaches work better
  on different stops.
• Measuring duration of stop state works reasonably
  well for /t,k/ b/c longer VOT than /p/.
• Detecting insertion of an aspiration model during
  decoding works well for /p,t/ but not k, which
  has too many false positives.
• Comparing phone probabilities worked well except
  for unaspirated /p/.

• VOT distinguishes consonants with the same place
  of articulation (/p/ vs. /b/, /t/ vs. /d/, etc.)
• However, different languages use different VOT
  intervals in contrasts (e.g. taco, pasta).
• English voiceless stops VOT 40-50 ms
• Spanish voiceless stops VOT near zero
• English voiced stops VOT near zero
• Spanish voiced stops negative VOT (voicing
  before closure

Future Work
Acknowledgements

• Since VOT is a time/timing related phenomenon, it
  may help to explicitly model the state duration
  density in the HMMs.
• Other optimisation criteria might be better
  suited than maximum likelihood estimation to
  train models for this purpose.
• More traditional classification approaches.

Special Thanks to the Tball Project for the data,
EE619 class for feedback, and Daylen Riggs and
Nathan Go for help with the transcriptions. Refer
ences available on request.