Morgan Sonderegger

Automatic measurement of voice onset time using discriminative structured predictiona)

A discriminative large-margin algorithm for automatic measurement of voice onset time (VOT) is described, considered as a case of predicting structured output from speech. Manually labeled data are used to train a function that takes as input a speech segment of an arbitrary length containing a voiceless stop, and outputs its VOT. The function is explicitly trained to minimize the difference between predicted and manually measured VOT; it operates on a set of acoustic feature functions designed based on spectral and temporal cues used by human VOT annotators. The algorithm is applied to initial voiceless stops from four corpora, representing different types of speech. Using several evaluation methods, the algorithms performance is near human intertranscriber reliability, and compares favorably with previous work. Furthermore, the algorithms performance is minimally affected by training and testing on different corpora, and remains essentially constant as the amount of training data is reduced to 50-250 manually labeled examples, demonstrating the methods practical applicability to new datasets.

Phonetic imitation from an individual-difference perspective: subjective attitude, personality and "autistic" traits.

Numerous studies have documented the phenomenon of phonetic imitation: the process by which the production patterns of an individual become more similar on some phonetic or acoustic dimension to those of her interlocutor. Though social factors have been suggested as a motivator for imitation, few studies has established a tight connection between language-external factors and a speaker’s likelihood to imitate. The present study investigated the phenomenon of phonetic imitation using a within-subject design embedded in an individual-differences framework. Participants were administered a phonetic imitation task, which included two speech production tasks separated by a perceptual learning task, and a battery of measures assessing traits associated with Autism-Spectrum Condition, working memory, and personality. To examine the effects of subjective attitude on phonetic imitation, participants were randomly assigned to four experimental conditions, where the perceived sexual orientation of the narrator (homosexual vs. heterosexual) and the outcome (positive vs. negative) of the story depicted in the exposure materials differed. The extent of phonetic imitation by an individual is significantly modulated by the story outcome, as well as by the participant’s subjective attitude toward the model talker, the participant’s personality trait of openness and the autistic-like trait associated with attention switching.

The private life of stops: VOT in a real-time corpus of spontaneous Glaswegian

Abstract While voice onset time (VOT) is known to be sensitive to a range of phonetic and linguistic factors, much less is known about VOT in spontaneous speech, since most studies consider stops in single words, in sentences, and/or in read speech. Scottish English is typically said to show less aspirated voiceless stops than other varieties of English, but there is also variation, ranging from unaspirated stops in vernacular speakers to more aspirated stops in Scottish Standard English; change in the vernacular has also been suggested. This paper presents results from a study which used a fast, semi-automated procedure for analyzing positive VOT, and applied it to stressed syllable-initial stops from a real- and apparent-time corpus of naturally-occurring spontaneous Glaswegian vernacular speech. We confirm significant effects on VOT for place of articulation and local speaking rate, and trends for vowel height and lexical frequency. With respect to time, our results are not consistent with previous work reporting generally shorter VOT in elderly speakers, since our results from models which control for local speech rate show lengthening over real-time in the elderly speakers in our sample. Overall, our findings suggest that VOT in both voiceless and voiced stops is lengthening over the course of the twentieth century in this variety of Scottish English. They also support observations from other studies, both from Scotland and beyond, indicating that gradient shifts along the VOT continuum reflect subtle sociolinguistic control.

Subglottal coupling and vowel space.

Poster presented at the 147th Meeting of the Acoustical Society of America, New York, N.Y., May 2004.

The emergence, progress, and impact of sound change in progress in Seoul Korean: Implications for mechanisms of tonogenesis

Abstract This study examines the origin, progression, and impact of a sound change in Seoul Korean where the primary cue to a stop contrast in phrase-initial position is shifting from VOT to f0. Because it shares similarities with the initial phase of tonogenesis, investigating this “quasi-tonogenetic” sound change provides insight into the nature of the emergence of contrastive f0 in “tonogenetic” sound changes more generally. Using a dataset from a large apparent-time corpus of Seoul Korean, we built mixed-effects regression models of VOT and f0 to examine the time-course of change, focusing on word frequency and vowel height effects. We found that both VOT contrast reduction and f0 contrast enhancement are more advanced in high-frequency words and in stops before non-high vowels, indicating that the change is spreading across words and phonetic contexts in parallel. Furthermore, speakers suppress non-contrastive variation in f0 as f0 emerges as a primary cue. Our findings suggest that one impetus for tonogenetic change is production bias coupled with an adaptive link between the cues. We further discuss the role of Korean intonational phonology on f0 which may help explain why the phonetic precondition leads to change in Seoul Korean but not in other languages.

Automatic forced alignment on child speech: Directions for improvement

Phonetic analysis is labor intensive, limiting the amount of data that can be considered. Recently, automated techniques (e.g., forced alignment based on Automatic Speech Recognition - ASR) have emerged allowing for much larger-scale analyses. For adult speech, forced alignment can be accurate even when the phonetic transcription is automatically generated, allowing for large-scale phonetic studies. However, such analyses remain difficult for childrens speech, where ASR methods perform more poorly. The present study used a trainable forced aligner that performs well on adult speech to examine the effect of four factors on alignment accuracy of child speech: (1) Corpus - elicited speech (multiple children) versus spontaneous speech (single child); (2) Pronunciation dictionary - standard adult versus customized; (3) Training data - adult lab speech, corpus-specific child speech, all child speech, or a combination of child and adult speech; (4) Segment type - voiceless stops, voiceless sibilants, and vowels...

Mixed-effects design analysis for experimental phonetics

Abstract It is common practice in the statistical analysis of phonetic data to draw conclusions on the basis of statistical significance. While p-values reflect the probability of incorrectly concluding a null effect is real, they do not provide information about other types of error that are also important for interpreting statistical results. In this paper, we focus on three measures related to these errors. The first, power, reflects the likelihood of detecting an effect that in fact exists. The second and third, Type M and Type S errors, measure the extent to which estimates of the magnitude and direction of an effect are inaccurate. We then provide an example of design analysis ( Gelman & Carlin, 2014 ), using data from an experimental study on German incomplete neutralization, to illustrate how power, magnitude, and sign errors vary with sample and effect size. This case study shows how the informativity of research findings can vary substantially in ways that are not always, or even usually, apparent on the basis of a p-value alone. We conclude by repeating three recommendations for good statistical practice in phonetics from best practices widely recommended for the social and behavioral sciences: report all results; design studies which will produce high-precision estimates; and conduct direct replications of previous findings.

The relationship of VOT and F0 contrasts across speakers and words in the German voicing contrast

Recent studies on tonogenesis in progress in Seoul Korean (Kang, 2014; Bang et al., 2015) find that the size of the VOT contrast and the f0 contrast between aspirated and lax stops “trade off” across speakers (e.g., male speakers have greater/smaller VOT/f0 contrasts), as well as words (e.g., different frequencies, following vowel heights). We examine whether this parallelism across speakers and words occurs in a language not undergoing tonogenesis by examining the size of the fortis/lenis contrast in VOT and f0 in German, using speech from the PhonDat corpus (Draxler, 1995). Mixed-effect regression models show that the size of the VOT contrast, but not the f0 contrast, is affected by properties of words (e.g., frequency, vowel height), unlike the parallelism observed in Korean. We further investigated whether VOT/f0 parallelism would be observed across speakers by partialing out linguistic factors affecting VOT/f0, and performing one logistic regression per speaker (n = 76) of fortis/lenis class as a fun...

Subglottal coupling as a quantal basis for the feature [back].

A model of acoustic coupling between the oral and subglottal cavities is described, which predicts attenuation of vowel formant prominences and discontinuities in formant trajectories near resonances of the subglottal system. The hypothesis that these effects on F2 near the second subglottal resonance (SubF2) are quantal effects for the feature [back] is examined using acoustic and subglottal data from English‐speaking adults. Experimental studies of F2 and SubF2 in English vowel production [Chi and Sonderegger, JASA 122, 1735–1745 (2007)] are reviewed and show that attenuation of second formant prominence and discontinuities in F2 trajectories near SubF2 consistently occur in back‐front diphthongs, in accordance with the acoustic model, while for monophthongs front and back vowel F2 values pattern above and below SubF2, as expected under the quantal hypothesis. An additional analysis of the data is presented, showing that breathiness, attenuation, and discontinuity are positively correlated across back‐f...