A Study on the Intelligibility of Korean-Accented English: Possibilities of Implementing AI Applications in English Education

Abstract

The current study investigates the generalizability of the Lingua Franca Core (LFC), conducting a Korean-accented English (KoE) intelligibility test with human subjects and with Artificial Intelligence mobile applications (AI apps). The data of KoE transcriptions collected from 30 native English speakers (NSs) and two AI apps were examined. The NSs group identified the words related to three LFC features significantly better than the AI apps, while with no significant difference, both groups recognized least frequently and with most difficulty the words that had non-LFC features. Despite NSs often performing better in the KoE intelligibility test, AI apps can capture speech realization at a level similar to the NSs (a moderate correlation found). The current results support the variability of both LFC and non-LFC feature sets and the novel application of the LFC analysis. Identifying difficult (less/least intelligible) English phonemes of KoE to NSs and to AI apps, we recommend that the variability of the LFC feature set should be taken into account in English language education and that learners should aim to communicate intelligibly with humans with various accents, by practicing/communicating with current available AI apps to make ready for the AI-Intervened/Mediated Circle.