Ask2Mask: Guided Data Selection for Masked Speech Modeling
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F22%3APU147442" target="_blank" >RIV/00216305:26230/22:PU147442 - isvavai.cz</a>
Result on the web
<a href="https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9806175" target="_blank" >https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9806175</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1109/JSTSP.2022.3186162" target="_blank" >10.1109/JSTSP.2022.3186162</a>
Alternative languages
Result language
angličtina
Original language name
Ask2Mask: Guided Data Selection for Masked Speech Modeling
Original language description
Masked speech modeling (MSM) methods such as wav2vec2 or w2v-BERT learn representations over speech frames which are randomlymaskedwithin an utterance. While thesemethods improve performance of Automatic Speech Recognition (ASR) systems, they have one major limitation. They treat all unsupervised speech samples with equal weight, which hinders learning as not all samples have relevant information to learn meaningful representations. In this work, we address this limitation. We propose ask2mask (ATM), a novel approach to focus on specific samples during MSM pre-training. ATM employs an external ASR model or scorer to weight unsupervised input samples in two different ways: 1) A fine-grained data selection is performed by masking over the highly confident input frames as chosen by the scorer. This allows themodel to learnmeaningful representations. 2) ATM is further extended to focus at utterance-level by weighting the final MSM loss with the utterance-level confidence score. We conduct fine-tuning experiments on two well-benchmarked corpora: LibriSpeech (matching the pre-training data) and Commonvoice, TED-LIUM, AMI and CHiME-6 (not matching the pre-training data). The results substantiate the efficacy of ATM on significantly improving the recognition performance under mismatched conditions (up to 11.6% relative over published results and upto 4.46% relative over our internal baseline) while still yielding modest improvements under matched conditions.
Czech name
—
Czech description
—
Classification
Type
J<sub>imp</sub> - Article in a specialist periodical, which is included in the Web of Science database
CEP classification
—
OECD FORD branch
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Result continuities
Project
—
Continuities
R - Projekt Ramcoveho programu EK
Others
Publication year
2022
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Name of the periodical
IEEE J-STSP
ISSN
1932-4553
e-ISSN
1941-0484
Volume of the periodical
16
Issue of the periodical within the volume
6
Country of publishing house
US - UNITED STATES
Number of pages
10
Pages from-to
1357-1366
UT code for WoS article
000870301500019
EID of the result in the Scopus database
2-s2.0-85133786585