A systematic literature review of research on automatic speech recognition in EFL pronunciation

To help EFL learners improve English pronunciation skills, researchers have recently turned to automatic speech recognition (ASR) technology. Despite some attempted review studies, relatively few have targeted ASR in an EFL setting. Therefore, the current study aims to delve into various aspects of ASR implementation, including theoretical foundations, research topics, research methods, target features, measurements, and findings, drawing upon published empirical literature. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guided the review. Of the 6138 articles with 2886 from Web of Science Core Collection and 3252 from Scopus, a total of 24 empirical articles were included. The results revealed: (1) twenty percent of the studies consisted of underpinning theories/models/concepts for the implementation of ASR in an EFL context; (2) a majority of studies adopted a quasi-experimental design, primarily focusing on the pronunciation gains in accuracy; (3) most learning activities with the use of ASR tools were designed for in-class activities; (4) the reviewed ASR studies were primarily concerned with the production accuracy of segmental features. Based on the findings, ASR is proven to be an effective tool in language classrooms but entails limitations. Future studies should investigate how ASR can best be implemented in a classroom setting.