Year: 2025
Type: article
Abstract: The presentation focuses on describing how automatic assessment of second/foreign language (L2) speech has advanced due to innovations in speech processing, machine learning, and natural language processing (NLP).Modern systems evaluate pronunciation, fluency, prosody, and intelligibility using a combination of acoustic, linguistic, and prosodic features (e.g.[3, 5]).Advances in artificial intelligence (AI) have led to improvements in assessment accuracy.These advancements have been integrated into commercial applications, including TOEFL®, Pearson's Versant, Duolingo English Test, and AI-driven tutoring systems like ELSA Speak.While English is still the most common language assessed automatically, significant developments are taking place also for many other languages (e.g.[1, 2, 4]).Despite progress, challenges remain in data scarcity, accent robustness, bias mitigation, and adaptive feedback.Current systems still struggle with diverse L2 accents and ensuring fairness in automated scoring.Future developments are likely to focus on multimodal integration (speech, facial expressions, gestures), explainable AI feedback, and personalized adaptive learning models to improve language learning experiences.
Source: Aaltodoc (Aalto University)
Authors: Mikko Kurimo, Tamás Grósz
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7023963288