Skip to content

Opening book details…

About this document

Contrastive Learning for Noisy ASR in SLU by zakari is a document available to read on EtoBox.

The paper introduces a two-stage method called Contrastive and Consistency Learning (CCL) to enhance neural noisy-channel models for Spoken Language Understanding (SLU) by addressing inconsistencies in Automatic Speech Recognition (ASR) transcripts. CCL correlates error patterns between clean and noisy transcripts and emphasizes the consistency of their latent features, improving intent classification performance in noisy environments. Experiments demonstrate that CCL outperforms existing methods across var

Author
zakari
Language
EN