Artificial intelligence in evaluating spoken English: Challenges and future perspectives

Abstract

In the context of accelerated globalization, the prominence of spoken English for cross-cultural communication has been underscored. A growing demand exists for objective, large-scale, and efficient evaluations of spoken English. Artificial intelligence (AI) has consequently been rigorously investigated and incorporated in spoken English evaluation. Nonetheless, much of the existing research tends to focus on singular evaluation dimensions, leading to performance limitations when confronted with diverse, stylistically varied, or cross-cultural English samples. In this study, multiple dimensions of spoken English evaluation are explored, and a novel evaluation model, utilizing adversarial training networks, is introduced. The aspiration is to elevate the accuracy and extensive adaptability of such evaluations.