How far can VOT take us? Voicing categorization with and without the use of VOT
- Abigail Benecke (Villanova University)
- Joseph Toscano (Villanova University)
Abstract
Voice-onset time (VOT) is an extremely reliable cue to word-initial stop voicing, such that VOT alone may be sufficient as a voicing cue. To test this, 35 potential cues were measured and used to train logistic regression classifiers, asking whether VOT is sufficient, whether other cues increase categorization accuracy, and whether, without VOT, other cues produce listener-level accuracy. Results show that human-like performance was never achieved without VOT or with VOT alone. Models using a cue-integration approach (additively combining multiple cues) offered the closest performance to human listeners. Thus, VOT appears to be necessary, but not sufficient, for voicing judgments.
Keywords: voicing, voice onset time, acoustic cues, phoneme categorization
How to Cite:
Benecke, A. & Toscano, J., (2018) “How far can VOT take us? Voicing categorization with and without the use of VOT”, Society for Computation in Linguistics 1(1), 205-206. doi: https://doi.org/10.7275/R5M043KH
Downloads:
Download PDF