The potential of Automatic Speech Recognition for fostering pronunciation learners\u27 autonomy

Despite ESL students frequently reporting a need or desire to work on their pronunciation in English, pronunciation is often downgraded as a teaching goal and often pushed aside in favor of other skills (Kelly, 1969; Lang, Wang, Shen, & Wang, 2012). Students that want to practice outside of class are likely to feel at a loss because they struggle to monitor their own speech and may not be able to get the feedback necessary to make improvements to their pronunciation. Students need skills, strategies, and resources that will allow them to work on their pronunciation on their own, less reliant on a teacher or school for pronunciation training. In effect, students need to learn to become autonomous learners of pronunciation. Automatic Speech Recognition (ASR) has great potential as a technology to help students get feedback on their pronunciation, allowing them to be more autonomous pronunciation learners. This study seeks to examine the effect of ASR on students\u27 autonomous learning beliefs and behaviors. Three groups, a control group (TRAD, n=15) which received traditional face-to-face (F2F) instruction, an experimental group (STRAT, n=17) which received traditional F2F instructions, but also minimal strategy training in ASR, and a second experimental group (HYBRID, n=16) which received hybrid instruction (half F2F with minimal strategy training and half working with ASR) were given a three-week pronunciation workshop on consonants and vowels of English known to be problematic for ESL students. Changes in beliefs of autonomy were measured through pre- and post-workshop surveys with Likert scale items as well as semi-structured interviews. Autonomous learning behaviors were monitored through self-reports of behavior during the course with language learning logs and after the course with a delayed post-workshop survey. Students explained choices to continue or stop working with ASR during a focus group at the end of the study. Results showed that STRAT and HYBRID both significantly increased their beliefs of autonomy from the pre- to post-workshop survey (for STRAT p=.006 and for HYBRID p=.013), while TRAD did not (p=.727). Students primarily pointed to ASR as the reason that they felt more capable of practicing their pronunciation on their own, stating that the ASR was useful for feedback because they could not hear their own errors when speaking. HYBRID reported significantly more time spent on autonomous pronunciation learning than STRAT and TRAD after the pronunciation workshop (p=.011). HYBRID also reported significantly more use of dictation software for pronunciation practice after the workshop than STRAT (p=.041)