Lecture: On Forced-Alignment and its Utility in Corpora for Phonetic Research

The VoxClamantis Corpus is a large corpus containing recordings of Bible readings, which have been forced aligned. The creators of the corpus promote it as a tool for large-scale typological research in phonetics. In this talk I first give some background on forced alignment and how it was applied to this corpus. I then I describe a small study I carried out which aimed to use the corpus to investigate which acoustic correlates differentiate the voiced aspirated affricate series from the voiced unaspirated affricate series in Bengali and Marathi. I discuss the benefits and limitations of using a forced aligned corpus for this type of research. Finally, I conclude that more work needs to be done to improve the performance of forced aligners before corpora such as the VoxClamantis Corpus can be used reliably for phonetics research, both for small-scale studies like mine, and for the large-scale typological research that the creators of the corpus intended.

Info

Day: 2022-05-26
Start time: 14:30
Duration: 00:30
Room: Decartes (2.36)
Track: Phonetics and Phonology
Language: en

Links:

Feedback

Click here to let us know how you liked this event.

Concurrent Events