Can I read Transformer Model for ASR Rescoring on EtoBox?

Transformer Model for ASR Rescoring by gogigorgonzola is a document available to read on EtoBox.

What is Transformer Model for ASR Rescoring about?

The document presents a novel Transformer-based model for Automatic Speech Recognition (ASR) that performs both rescoring and rewriting of N-best hypotheses, enhancing the efficiency and accuracy of voice assistants. The proposed model, known as the Transformer Rescore Attention (TRA), utilizes a new training objective called Matching Query Similarity Distribution (MQSD) to improve performance, achieving an average 8.6% reduction in Word Error Rate (WER) compared to existing methods. This approach maintains

Author

gogigorgonzola

Language

EN