Speech Scoring

Automatic speech scoring (ASS) aims to objectively assess various aspects of spoken language, from voice quality and pronunciation to overall language proficiency, using computational methods. Current research emphasizes improving ASS accuracy and interpretability through techniques like deep learning (e.g., transformer networks and convolutional neural networks), hybrid models combining handcrafted and self-supervised features, and addressing data imbalance issues with novel regularization methods. These advancements hold significant promise for applications in healthcare (e.g., detecting voice disorders), language education (e.g., providing personalized feedback), and clinical practice (e.g., standardizing voice quality assessments).