Crowd-sourcing a benchmark for English forced alignment

Michael McAuliffe (McGill)
Kaylynn Gunter (U. of Oregon)

Description: Do you use forced aligned English data? Do you wish forced aligners were more principled in placing segmental boundaries? If so, this workshop is for you! We are crowd-sourcing a manually-corrected benchmark for English forced aligners that can be easily distributed and used in the development of new models and architectures. The workshop will be a half day of correcting forced aligned English read speech according to formalized conventions, provided by the workshop organizers. So, just bring your laptop, your interest in fine phonetic detail, and your desire to create more linguistically-informed tools and open data sets!