Volume 81, Issue 5

Given an orthographic transcription, forced alignment systems automatically determine boundaries between segments in speech, facilitating the use of large corpora. In the present paper, we introduce a neural network-based forced alignment system, the Mason-Alberta Phonetic Segmenter (MAPS). MAPS serves as a testbed for two possible improvements we pursue for forced alignment systems. The first is treating the acoustic model as a tagger, rather than a classifier, motivated by the common understanding that segments are not truly discrete and often overlap. The second is an interpolation technique to allow more precise boundaries than the typical 10 ms limit in modern systems. During testing, all system configurations we trained significantly outperformed the state-of-the-art Montreal Forced Aligner in the 10 ms boundary placement tolerance threshold. The greatest difference achieved was a 28.13 % relative performance increase. The Montreal Forced Aligner began to slightly outperform our models at around a 30 ms tolerance. We also reflect on the training process for acoustic modeling in forced alignment, highlighting how the output targets for these models do not match phoneticians’ conception of similarity between phones and that reconciling this tension may require rethinking the task and output targets or how speech itself should be segmented.

Rhotic sounds are well known for their considerable phonetic variation within and across languages and their complexity in speech production. Although rhotics in many languages have been examined and documented, the phonetic features of Mandarin rhotics remain unclear, and debates about the prevocalic rhotic (the syllable-onset rhotic) persist. This paper extends the investigation of rhotic sounds by examining the articulatory and acoustic features of Mandarin Chinese rhotics in prevocalic, syllabic (the rhotacized vowel [ɚ]), and postvocalic (r-suffix) positions. Eighteen speakers from Northern China were recorded using ultrasound imaging. Results showed that Mandarin syllabic and postvocalic rhotics can be articulated with various tongue shapes, including tongue-tip-up retroflex and tongue-tip-down bunched shapes. Different tongue shapes have no significant acoustic differences in the first three formants, demonstrating a many-to-one articulation-acoustics relationship. The prevocalic rhotics in our data were found to be articulated only with bunched tongue shapes, and were sometimes produced with frication noise at the start. In general, rhotics in all syllable positions are characterized by a close F2 and F3, though the prevocalic rhotic has a higher F2 and F3 than the syllabic and postvocalic rhotics. The effects of syllable position and vowel context are also discussed.