'I Can Speak': Improving English Pronunciation through Automatic Speech Recognition-Based Language Learning Systems

Practicing pronunciation through language learning systems incorporating Automatic Speech Recognition (ASR) technology has been effective in helping improve foreign language pronunciation. One of the ASR affordances is that it can provide immediate, personalized feedback on learners' pronunciation. We investigated the effects of two ASR-based language learning systems, "I Love Indonesia" (ILI) and "NovoLearning" (NOVO), on learners' word-level and sentence-level pronunciation. ILI offers global corrective feedback, while NOVO is equipped with corrective feedback on phonetic details. 117 Indonesian high school students participated in a five-week-long experiment; 52 students used ILI and 65 NOVO. Three pronunciation measures were calculated on a pre and post reading test: phonetic edit distances plus accentedness and comprehensibility ratings. Results indicate significant improvements in learners' pronunciation, confirming that both systems are promising learning tools, with NOVO leading to more progress. Future studies should examine the long-term effects of ASR-based global and phonetic feedback on learners' pronunciation quality.