A presentation by Carol Johnson, PhD student Concordia University, lecturer, École de langues. 

This study examined the use of a free and popular automatic speech recognition (ASR) technology, Google voice typing (GVT), to automatically assess English as second language pronunciation for a university’s English placement test. It aimed to answer the following question: What is the relationship between GVT-rated scores and human-rated scores? To answer this question, we compared GVT and human-rated scores of audio recordings from fifty-six undergraduate students’ placement tests at a university in Canada, whose oral production levels ranged from A1 to C2. Using a rubric that assessed comprehensibility, phoneme accuracy, and prosody (connected speech, word/sentence stress, and intonation), three experienced ESL instructors came to a consensus about each participant’s score. The same recordings were then played into GVT. The output was analyzed to determine the number of correctly recognized words, and a proportional score was generated. Our results indicate that GVT scores strongly correlated with human-rater scores, indicating that this free ASR technology could be leveraged to increase the test usefulness of language placement tests. 

Bimodal activity. Those who wish to attend via Zoom will receive a link with the confirmation email after RSVPing. 

RSVP: https://uqam-ca.libcal.com/calendar/melse?cid=8040&t=d&d=0000-00-00&cal=8040&inc=0 

Info: springer.suzanne@uqam.ca