AI That Listens and Reads: Multimodal Deep Learning Scores Spoken English With 97.5% Accuracy
A new multimodal deep learning framework fuses Log-Mel spectrograms and BERT embeddings through self-attention to automatically score spoken English fluency ...
