Traditional Versus ASR-Based Pronunciation Instruction

Abstract

This paper presents a 15-week classroom study measuring the student outcomes of instructor-led pronunciation lessons versus entirely ASR-based pronunciation training. Seventy-six second-semester Spanish language learners were divided into two groups, one experimental (n=44) and one control (n=32). Over the course of six modules, both groups completed a pre- and post-study recording, as well as explicit pronunciation training sessions. These sessions included pre- and post-recordings, with either traditional or ASR pronunciation practice in between, which aimed attention at targeted phonemes. All student recordings were evaluated by native and near-natives for comprehensibility, nativeness, fluency, and perceived confidence. The results show that the effect of explicit and ASR instruction varies depending on the module and characteristic evaluated. ASR seems to outperform traditional instruction when targeting specific phonemes, especially in the short-term, while the explicit instruction group saw longer-term gains in regards to comprehensibility. Holistically, the data suggest that ASR-based instruction shows promise to improve certain aspects of pronunciation, but that using both techniques in tandem would be the most strategic approach to handling the development of this fundamental aspect of learner speech. The data presented here highlight the role and effectiveness of computer-assisted pronunciation training for lower-level Spanish courses.