Locality enhanced dynamic biasing and sampling strategies for contextual ASR

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jalal, Md Asif, Parada, Pablo Peso, Pavlidis, George, Moschopoulos, Vasileios, Saravanan, Karthikeyan, Kontoulis, Chrysovalantis-Giorgos, Zhang, Jisi, Drosou, Anastasios, Lee, Gil Ho, Lee, Jungin, Jung, Seokyeong
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929221236948992
author Jalal, Md Asif
Parada, Pablo Peso
Pavlidis, George
Moschopoulos, Vasileios
Saravanan, Karthikeyan
Kontoulis, Chrysovalantis-Giorgos
Zhang, Jisi
Drosou, Anastasios
Lee, Gil Ho
Lee, Jungin
Jung, Seokyeong
author_facet Jalal, Md Asif
Parada, Pablo Peso
Pavlidis, George
Moschopoulos, Vasileios
Saravanan, Karthikeyan
Kontoulis, Chrysovalantis-Giorgos
Zhang, Jisi
Drosou, Anastasios
Lee, Gil Ho
Lee, Jungin
Jung, Seokyeong
contents Automatic Speech Recognition (ASR) still face challenges when recognizing time-variant rare-phrases. Contextual biasing (CB) modules bias ASR model towards such contextually-relevant phrases. During training, a list of biasing phrases are selected from a large pool of phrases following a sampling strategy. In this work we firstly analyse different sampling strategies to provide insights into the training of CB for ASR with correlation plots between the bias embeddings among various training stages. Secondly, we introduce a neighbourhood attention (NA) that localizes self attention (SA) to the nearest neighbouring frames to further refine the CB output. The results show that this proposed approach provides on average a 25.84% relative WER improvement on LibriSpeech sets and rare-word evaluation compared to the baseline.
format Preprint
id arxiv_https___arxiv_org_abs_2401_13146
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Locality enhanced dynamic biasing and sampling strategies for contextual ASR
Jalal, Md Asif
Parada, Pablo Peso
Pavlidis, George
Moschopoulos, Vasileios
Saravanan, Karthikeyan
Kontoulis, Chrysovalantis-Giorgos
Zhang, Jisi
Drosou, Anastasios
Lee, Gil Ho
Lee, Jungin
Jung, Seokyeong
Audio and Speech Processing
Computation and Language
Sound
Automatic Speech Recognition (ASR) still face challenges when recognizing time-variant rare-phrases. Contextual biasing (CB) modules bias ASR model towards such contextually-relevant phrases. During training, a list of biasing phrases are selected from a large pool of phrases following a sampling strategy. In this work we firstly analyse different sampling strategies to provide insights into the training of CB for ASR with correlation plots between the bias embeddings among various training stages. Secondly, we introduce a neighbourhood attention (NA) that localizes self attention (SA) to the nearest neighbouring frames to further refine the CB output. The results show that this proposed approach provides on average a 25.84% relative WER improvement on LibriSpeech sets and rare-word evaluation compared to the baseline.
title Locality enhanced dynamic biasing and sampling strategies for contextual ASR
topic Audio and Speech Processing
Computation and Language
Sound
url https://arxiv.org/abs/2401.13146