FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3160799772> ?p ?o ?g. }

• W3160799772 abstract "Compared to vision and language applications, self-supervised pre-training approaches for ASR are challenged by three unique problems: (1) There are multiple sound units in each input utterance, (2) With audio-only pre-training, there is no lexicon of sound units, and (3) Sound units have variable lengths with no explicit segmentation. In this paper, we propose the Hidden-Unit BERT (HUBERT) model which utilizes a cheap k-means clustering step to provide aligned target labels for pre-training of a BERT model. A key ingredient of our approach is applying the predictive loss over the masked regions only. This allows the pre-training stage to benefit from the consistency of the unsupervised teacher rather that its intrinsic quality. Starting with a simple k-means teacher of 100 cluster, and using two iterations of clustering, the HUBERT model matches the state-of-the-art wav2vec 2.0 performance on the ultra low-resource Libri-light 10h, 1h, 10min supervised subsets." @default.
• W3160799772 created "2021-05-24" @default.
• W3160799772 creator A5015826285 @default.
• W3160799772 creator A5042186348 @default.
• W3160799772 creator A5044673988 @default.
• W3160799772 creator A5051950818 @default.
• W3160799772 creator A5071983998 @default.
• W3160799772 date "2021-06-06" @default.
• W3160799772 modified "2023-10-06" @default.
• W3160799772 title "Hubert: How Much Can a Bad Teacher Benefit ASR Pre-Training?" @default.
• W3160799772 cites W1494198834 @default.
• W3160799772 cites W2110073835 @default.
• W3160799772 cites W2127141656 @default.
• W3160799772 cites W2146444479 @default.
• W3160799772 cites W2750248772 @default.
• W3160799772 cites W2933138175 @default.
• W3160799772 cites W2946417913 @default.
• W3160799772 cites W2962739339 @default.
• W3160799772 cites W2962850167 @default.
• W3160799772 cites W2973049979 @default.
• W3160799772 cites W2981283774 @default.
• W3160799772 cites W2982223350 @default.
• W3160799772 cites W2995181338 @default.
• W3160799772 cites W3003875258 @default.
• W3160799772 cites W3008525923 @default.
• W3160799772 cites W3011411500 @default.
• W3160799772 cites W3015265920 @default.
• W3160799772 cites W3015522062 @default.
• W3160799772 cites W3035202887 @default.
• W3160799772 cites W3035524453 @default.
• W3160799772 cites W3048217718 @default.
• W3160799772 cites W3125709657 @default.
• W3160799772 cites W4254197176 @default.
• W3160799772 doi "https://doi.org/10.1109/icassp39728.2021.9414460" @default.
• W3160799772 hasPublicationYear "2021" @default.
• W3160799772 type Work @default.
• W3160799772 sameAs 3160799772 @default.
• W3160799772 citedByCount "40" @default.
• W3160799772 countsByYear W31607997722021 @default.
• W3160799772 countsByYear W31607997722022 @default.
• W3160799772 countsByYear W31607997722023 @default.
• W3160799772 crossrefType "proceedings-article" @default.
• W3160799772 hasAuthorship W3160799772A5015826285 @default.
• W3160799772 hasAuthorship W3160799772A5042186348 @default.
• W3160799772 hasAuthorship W3160799772A5044673988 @default.
• W3160799772 hasAuthorship W3160799772A5051950818 @default.
• W3160799772 hasAuthorship W3160799772A5071983998 @default.
• W3160799772 hasConcept C115961682 @default.
• W3160799772 hasConcept C119857082 @default.
• W3160799772 hasConcept C121332964 @default.
• W3160799772 hasConcept C137293760 @default.
• W3160799772 hasConcept C153294291 @default.
• W3160799772 hasConcept C154945302 @default.
• W3160799772 hasConcept C155635449 @default.
• W3160799772 hasConcept C204321447 @default.
• W3160799772 hasConcept C26517878 @default.
• W3160799772 hasConcept C2775852435 @default.
• W3160799772 hasConcept C2776436953 @default.
• W3160799772 hasConcept C2777211547 @default.
• W3160799772 hasConcept C2778121359 @default.
• W3160799772 hasConcept C28490314 @default.
• W3160799772 hasConcept C38652104 @default.
• W3160799772 hasConcept C41008148 @default.
• W3160799772 hasConcept C61328038 @default.
• W3160799772 hasConcept C73555534 @default.
• W3160799772 hasConcept C89600930 @default.
• W3160799772 hasConcept C99498987 @default.
• W3160799772 hasConceptScore W3160799772C115961682 @default.
• W3160799772 hasConceptScore W3160799772C119857082 @default.
• W3160799772 hasConceptScore W3160799772C121332964 @default.
• W3160799772 hasConceptScore W3160799772C137293760 @default.
• W3160799772 hasConceptScore W3160799772C153294291 @default.
• W3160799772 hasConceptScore W3160799772C154945302 @default.
• W3160799772 hasConceptScore W3160799772C155635449 @default.
• W3160799772 hasConceptScore W3160799772C204321447 @default.
• W3160799772 hasConceptScore W3160799772C26517878 @default.
• W3160799772 hasConceptScore W3160799772C2775852435 @default.
• W3160799772 hasConceptScore W3160799772C2776436953 @default.
• W3160799772 hasConceptScore W3160799772C2777211547 @default.
• W3160799772 hasConceptScore W3160799772C2778121359 @default.
• W3160799772 hasConceptScore W3160799772C28490314 @default.
• W3160799772 hasConceptScore W3160799772C38652104 @default.
• W3160799772 hasConceptScore W3160799772C41008148 @default.
• W3160799772 hasConceptScore W3160799772C61328038 @default.
• W3160799772 hasConceptScore W3160799772C73555534 @default.
• W3160799772 hasConceptScore W3160799772C89600930 @default.
• W3160799772 hasConceptScore W3160799772C99498987 @default.
• W3160799772 hasLocation W31607997721 @default.
• W3160799772 hasOpenAccess W3160799772 @default.
• W3160799772 hasPrimaryLocation W31607997721 @default.
• W3160799772 hasRelatedWork W2066913438 @default.
• W3160799772 hasRelatedWork W2110052520 @default.
• W3160799772 hasRelatedWork W2128201184 @default.
• W3160799772 hasRelatedWork W2165348108 @default.
• W3160799772 hasRelatedWork W2374918184 @default.
• W3160799772 hasRelatedWork W2399356099 @default.
• W3160799772 hasRelatedWork W2612720144 @default.
• W3160799772 hasRelatedWork W2955724459 @default.
• W3160799772 hasRelatedWork W4221138681 @default.
• W3160799772 hasRelatedWork W2972741966 @default.

• next