Matches in SemOpenAlex for { <https://semopenalex.org/work/W3144810982> ?p ?o ?g. }
Showing items 1 to 79 of
79
with 100 items per page.
- W3144810982 abstract "Contrastive Predictive Coding (CPC), based on predicting future segments of speech from past segments is emerging as a powerful algorithm for representation learning of speech signal. However, it still under-performs compared to other methods on unsupervised evaluation benchmarks. Here, we intro-duce WavAugment, a time-domain data augmentation library which we adapt and optimize for the specificities of CPC (raw waveform input, contrastive loss, past versus future structure). We find that applying augmentation only to the segments from which the CPC prediction is performed yields better results than applying it also to future segments from which the samples (both positive and negative) of the contrastive loss are drawn. After selecting the best combination of pitch modification, additive noise and reverberation on unsupervised metrics on LibriSpeech (with a gain of 18-22% relative on the ABX score), we apply this combination without any change to three new datasets in the Zero Resource Speech Benchmark 2017 and beat the state-of-the-art using out-of-domain training data. Finally, we show that the data-augmented pretrained features improve a downstream phone recognition task in the Libri-light semi-supervised setting (10 min, 1 h or 10 h of labelled data) reducing the PER by 15% relative." @default.
- W3144810982 created "2021-04-13" @default.
- W3144810982 creator A5007620149 @default.
- W3144810982 creator A5034757620 @default.
- W3144810982 creator A5040376974 @default.
- W3144810982 creator A5041907084 @default.
- W3144810982 creator A5053822883 @default.
- W3144810982 creator A5060751140 @default.
- W3144810982 creator A5078102229 @default.
- W3144810982 date "2021-01-19" @default.
- W3144810982 modified "2023-09-30" @default.
- W3144810982 title "Data Augmenting Contrastive Learning of Speech Representations in the Time Domain" @default.
- W3144810982 cites W1494198834 @default.
- W3144810982 cites W1989674786 @default.
- W3144810982 cites W2020607164 @default.
- W3144810982 cites W2055408826 @default.
- W3144810982 cites W2346964103 @default.
- W3144810982 cites W2347098582 @default.
- W3144810982 cites W2509930204 @default.
- W3144810982 cites W2696967604 @default.
- W3144810982 cites W2883725317 @default.
- W3144810982 cites W2936774411 @default.
- W3144810982 cites W2973026522 @default.
- W3144810982 cites W3003875258 @default.
- W3144810982 cites W3015783745 @default.
- W3144810982 cites W3093096176 @default.
- W3144810982 cites W3125709657 @default.
- W3144810982 doi "https://doi.org/10.1109/slt48900.2021.9383605" @default.
- W3144810982 hasPublicationYear "2021" @default.
- W3144810982 type Work @default.
- W3144810982 sameAs 3144810982 @default.
- W3144810982 citedByCount "24" @default.
- W3144810982 countsByYear W31448109822021 @default.
- W3144810982 countsByYear W31448109822022 @default.
- W3144810982 countsByYear W31448109822023 @default.
- W3144810982 crossrefType "proceedings-article" @default.
- W3144810982 hasAuthorship W3144810982A5007620149 @default.
- W3144810982 hasAuthorship W3144810982A5034757620 @default.
- W3144810982 hasAuthorship W3144810982A5040376974 @default.
- W3144810982 hasAuthorship W3144810982A5041907084 @default.
- W3144810982 hasAuthorship W3144810982A5053822883 @default.
- W3144810982 hasAuthorship W3144810982A5060751140 @default.
- W3144810982 hasAuthorship W3144810982A5078102229 @default.
- W3144810982 hasBestOaLocation W31448109822 @default.
- W3144810982 hasConcept C13280743 @default.
- W3144810982 hasConcept C153180895 @default.
- W3144810982 hasConcept C154945302 @default.
- W3144810982 hasConcept C185798385 @default.
- W3144810982 hasConcept C205649164 @default.
- W3144810982 hasConcept C28490314 @default.
- W3144810982 hasConcept C41008148 @default.
- W3144810982 hasConceptScore W3144810982C13280743 @default.
- W3144810982 hasConceptScore W3144810982C153180895 @default.
- W3144810982 hasConceptScore W3144810982C154945302 @default.
- W3144810982 hasConceptScore W3144810982C185798385 @default.
- W3144810982 hasConceptScore W3144810982C205649164 @default.
- W3144810982 hasConceptScore W3144810982C28490314 @default.
- W3144810982 hasConceptScore W3144810982C41008148 @default.
- W3144810982 hasLocation W31448109821 @default.
- W3144810982 hasLocation W31448109822 @default.
- W3144810982 hasLocation W31448109823 @default.
- W3144810982 hasLocation W31448109824 @default.
- W3144810982 hasLocation W31448109825 @default.
- W3144810982 hasOpenAccess W3144810982 @default.
- W3144810982 hasPrimaryLocation W31448109821 @default.
- W3144810982 hasRelatedWork W112744582 @default.
- W3144810982 hasRelatedWork W1485630101 @default.
- W3144810982 hasRelatedWork W1490303524 @default.
- W3144810982 hasRelatedWork W1537221325 @default.
- W3144810982 hasRelatedWork W2030059621 @default.
- W3144810982 hasRelatedWork W2070338563 @default.
- W3144810982 hasRelatedWork W2164513229 @default.
- W3144810982 hasRelatedWork W2498017833 @default.
- W3144810982 hasRelatedWork W3033750096 @default.
- W3144810982 hasRelatedWork W3081841992 @default.
- W3144810982 isParatext "false" @default.
- W3144810982 isRetracted "false" @default.
- W3144810982 magId "3144810982" @default.
- W3144810982 workType "article" @default.