Advancements and Challenges in Automated Evaluation of Spoken Language Proficiency Using AI
Ebrahim Mohammadkarimi, Jamal Ali Omar, Abdulla Salim Rashid · Advances in computational intelligence and robotics book series · 2025
This research aims to examine the advancements and challenges in the automated evaluation of spoken skills through the use of artificial intelligence. We triangulate data collection using a Likert-scale questionnaire and open-ended interviews. In language assessment, the main participants were 79 students (38 elementary and 41 intermediate) and 21 experienced teachers (raters). Two AI-powered tools, Versant and Speechace, as well as raters, separately assessed the two speaking tests that students participated in. AI-powered tools showed that the correlation coefficients for grammar, fluency, vocabulary, and pronunciation parts are high. This means that the Versant and Speechace tests are strongly connected in a straight line. The outcomes of the questionnaires and interviews indicated that raters generally hold positive perceptions of AI's role in language proficiency assessment, although they noted some issues with AI`s understanding speaking, especially in detecting varieties of accents and dialects.