Implementing contextual biasing in GPU decoder for online ASR

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F23%3APU150719" target="_blank" >RIV/00216305:26230/23:PU150719 - isvavai.cz</a>

Result on the web

<a href="https://www.isca-archive.org/interspeech_2023/nigmatulina23_interspeech.html" target="_blank" >https://www.isca-archive.org/interspeech_2023/nigmatulina23_interspeech.html</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.21437/Interspeech.2023-2449" target="_blank" >10.21437/Interspeech.2023-2449</a>

Result language

angličtina

Original language name

Implementing contextual biasing in GPU decoder for online ASR

Original language description

GPU decoding significantly accelerates the output of ASR predictions. While GPUs are already being used for online ASR decoding, post-processing and rescoring on GPUs have not been properly investigated yet. Rescoring with available contextual information can considerably improve ASR predictions. Previous studies have proven the viability of lattice rescoring in decoding and biasing language model (LM) weights in offline and online CPU scenarios. In real-time GPU decoding, partial recognition hypotheses are produced without lattice generation, which makes the implementation of biasing more complex. The paper proposes and describes an approach to integrate contextual biasing in real-time GPU decoding while exploiting the standard Kaldi GPU decoder. Besides the biasing of partial ASR predictions, our approach also permits dynamic context switching allowing a flexible rescoring per each speech segment directly on GPU. The code is publicly released1 and tested with open-sourced test sets.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Project

—

Continuities

S - Specificky vyzkum na vysokych skolach

Publication year

2023

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

Proceedings of the Annual Conference of International Speech Communication Association, INTERSPEECH

ISBN

—

ISSN

1990-9772

e-ISSN

—

Number of pages

5

Pages from-to

4494-4498

Publisher name

International Speech Communication Association

Place of publication

Dublin

Event location

Dublin

Event date

Aug 20, 2023

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

—