Effectiveness of AI-Powered Speech Recognition App in Enhancing Speaking Skills of Intermediate-Level ELLs in Mianwali

Abstract

Traditional means of learning English, such as the Grammar Translation Method (GTM), have focused on the development of writing skills while neglecting speaking skills. This negligence has resulted in poor pronunciation, low proficiency, and speaking anxiety among English Language Learners (ELLs). In order to address this gap, the effectiveness of an AI-powered Automatic Speech Recognition ASR application, ELSA Speak, was assessed in this study among the intermediate-level students in the Mianwali district. 60 male students, who were recruited via purposive sampling at Government. Graduate College, Mianwali, were randomly assigned to either the Experimental Group, who used the designated application for 6 weeks, or the Control Group, who were taught traditionally. Their speaking performance was assessed through pre-tests and post-tests via a rubric. Semi-structured interviews were also conducted with 8 participants to explore their experience of and challenges regarding using this application and their perspectives were weighed through the Technology Acceptance Model. The quantitative analysis of the post-tests showed that EG outperformed the CG with mean Composite Scores of 19.33 and 14.90 respectively. The analysis of the interviews also yielded high perceived usefulness due to the application providing a non-judgemental environment that propelled learner autonomy, reduced their speaking anxiety, and boosted their confidence. However, some learners were also quick to point out challenges such as unreliable internet connections that made this application hard to use and the increased cost of mobile data. This study concludes that AI-powered ASR applications are an effective intervention to address the lack of attention given towards the development of speaking skills in overcrowded ESL classrooms.