Matches in SemOpenAlex for { <https://semopenalex.org/work/W4319862247> ?p ?o ?g. }
Showing items 1 to 100 of
100
with 100 items per page.
- W4319862247 abstract "Data augmentation is a ubiquitous technique used to provide robustness to automatic speech recognition (ASR) training. However, even as so much of the ASR training process has become automated and more “end-to-end,” the data augmentation policy (what augmentation functions to use, and how to apply them) remains hand-crafted. We present G(raph)-Augment, a technique to define the augmentation space as directed acyclic graphs (DAGs) and search over this space to optimize the augmentation policy itself. We show that given the same computational budget, policies produced by G-Augment are able to perform better than SpecAugment policies obtained by random search on fine-tuning tasks on CHiME-6 and AMI. G-Augment is also able to establish a new state-of-the-art ASR performance on the CHiME-6 evaluation set (30.7% WER). We further demonstrate that G- Augment policies show better transfer properties across warm-start to cold-start training and model size compared to random-searched SpecAugment policies." @default.
- W4319862247 created "2023-02-11" @default.
- W4319862247 creator A5007069562 @default.
- W4319862247 creator A5025935520 @default.
- W4319862247 creator A5031341964 @default.
- W4319862247 creator A5038305187 @default.
- W4319862247 creator A5039108749 @default.
- W4319862247 creator A5065360819 @default.
- W4319862247 creator A5071715737 @default.
- W4319862247 creator A5088551093 @default.
- W4319862247 creator A5090030865 @default.
- W4319862247 date "2023-01-09" @default.
- W4319862247 modified "2023-09-27" @default.
- W4319862247 title "G-Augment: Searching for the Meta-Structure of Data Augmentation Policies for ASR" @default.
- W4319862247 cites W1494198834 @default.
- W4319862247 cites W1659842140 @default.
- W4319862247 cites W1979651826 @default.
- W4319862247 cites W2020399841 @default.
- W4319862247 cites W2040492000 @default.
- W4319862247 cites W2111935653 @default.
- W4319862247 cites W2112796928 @default.
- W4319862247 cites W2136439176 @default.
- W4319862247 cites W2156163116 @default.
- W4319862247 cites W2184343439 @default.
- W4319862247 cites W2291975472 @default.
- W4319862247 cites W2407080277 @default.
- W4319862247 cites W2604262106 @default.
- W4319862247 cites W2696967604 @default.
- W4319862247 cites W2732547613 @default.
- W4319862247 cites W2954996726 @default.
- W4319862247 cites W2964081807 @default.
- W4319862247 cites W2965658867 @default.
- W4319862247 cites W2995181338 @default.
- W4319862247 cites W3015995734 @default.
- W4319862247 cites W3035160371 @default.
- W4319862247 cites W3095687973 @default.
- W4319862247 cites W3097777922 @default.
- W4319862247 cites W3160259745 @default.
- W4319862247 cites W3204696009 @default.
- W4319862247 cites W84880661 @default.
- W4319862247 doi "https://doi.org/10.1109/slt54892.2023.10022748" @default.
- W4319862247 hasPublicationYear "2023" @default.
- W4319862247 type Work @default.
- W4319862247 citedByCount "0" @default.
- W4319862247 crossrefType "proceedings-article" @default.
- W4319862247 hasAuthorship W4319862247A5007069562 @default.
- W4319862247 hasAuthorship W4319862247A5025935520 @default.
- W4319862247 hasAuthorship W4319862247A5031341964 @default.
- W4319862247 hasAuthorship W4319862247A5038305187 @default.
- W4319862247 hasAuthorship W4319862247A5039108749 @default.
- W4319862247 hasAuthorship W4319862247A5065360819 @default.
- W4319862247 hasAuthorship W4319862247A5071715737 @default.
- W4319862247 hasAuthorship W4319862247A5088551093 @default.
- W4319862247 hasAuthorship W4319862247A5090030865 @default.
- W4319862247 hasConcept C104317684 @default.
- W4319862247 hasConcept C11413529 @default.
- W4319862247 hasConcept C138885662 @default.
- W4319862247 hasConcept C154945302 @default.
- W4319862247 hasConcept C177264268 @default.
- W4319862247 hasConcept C185592680 @default.
- W4319862247 hasConcept C199360897 @default.
- W4319862247 hasConcept C2779070825 @default.
- W4319862247 hasConcept C41008148 @default.
- W4319862247 hasConcept C41895202 @default.
- W4319862247 hasConcept C51632099 @default.
- W4319862247 hasConcept C55493867 @default.
- W4319862247 hasConcept C63479239 @default.
- W4319862247 hasConcept C74197172 @default.
- W4319862247 hasConcept C98045186 @default.
- W4319862247 hasConceptScore W4319862247C104317684 @default.
- W4319862247 hasConceptScore W4319862247C11413529 @default.
- W4319862247 hasConceptScore W4319862247C138885662 @default.
- W4319862247 hasConceptScore W4319862247C154945302 @default.
- W4319862247 hasConceptScore W4319862247C177264268 @default.
- W4319862247 hasConceptScore W4319862247C185592680 @default.
- W4319862247 hasConceptScore W4319862247C199360897 @default.
- W4319862247 hasConceptScore W4319862247C2779070825 @default.
- W4319862247 hasConceptScore W4319862247C41008148 @default.
- W4319862247 hasConceptScore W4319862247C41895202 @default.
- W4319862247 hasConceptScore W4319862247C51632099 @default.
- W4319862247 hasConceptScore W4319862247C55493867 @default.
- W4319862247 hasConceptScore W4319862247C63479239 @default.
- W4319862247 hasConceptScore W4319862247C74197172 @default.
- W4319862247 hasConceptScore W4319862247C98045186 @default.
- W4319862247 hasLocation W43198622471 @default.
- W4319862247 hasOpenAccess W4319862247 @default.
- W4319862247 hasPrimaryLocation W43198622471 @default.
- W4319862247 hasRelatedWork W1480922982 @default.
- W4319862247 hasRelatedWork W1989218596 @default.
- W4319862247 hasRelatedWork W2355215981 @default.
- W4319862247 hasRelatedWork W2370652759 @default.
- W4319862247 hasRelatedWork W2584455473 @default.
- W4319862247 hasRelatedWork W4239295757 @default.
- W4319862247 hasRelatedWork W4303857474 @default.
- W4319862247 hasRelatedWork W4307074309 @default.
- W4319862247 hasRelatedWork W4319862247 @default.
- W4319862247 hasRelatedWork W831794578 @default.
- W4319862247 isParatext "false" @default.
- W4319862247 isRetracted "false" @default.
- W4319862247 workType "article" @default.