Aggregating the evidence of automatic speech recognition research claims in CALL

This review provides an overview and analysis of ASR (automatic speech recognition) research claims identified in CALL (computer-assisted language learning) studies from the past decade. When taken separately, few conclusions can be drawn and little extrapolation is possible from any one isolated investigation. Empirical studies on ASR vary tremendously in size, scope, and research questions posited. However, clear patterns and implications for educators and CALL researchers are beginning to emerge from the data. This research synthesis of 50 studies considers how effective ASR tools are at assessing and ultimately improving L2 pronunciation. Two novel rubrics are proposed to categorize evidence strength for individual studies and to further classify the frequency of claims across multiple studies. Results from this analysis suggest that there are strong empirical arguments in favor of utilizing ASR for pronunciation instruction and evaluation. Specifically, ASR has a demonstrated capacity to accurately identify errors, improve pronunciation, and is strongly associated with a positive student experience.