Matches in SemOpenAlex for { <https://semopenalex.org/work/W4310923040> ?p ?o ?g. }
Showing items 1 to 80 of
80
with 100 items per page.
- W4310923040 abstract "Self-supervised learning (SSL) models have achieved considerable improvements in automatic speech recognition (ASR). In addition, ASR performance could be further improved if the model is dedicated to audio content information learning theoretically. To this end, we propose a progressive multi-scale self-supervised learning (PMS-SSL) method, which uses fine-grained target sets to compute SSL loss at top layer while uses coarse-grained target sets at intermediate layers. Furthermore, PMS-SSL introduces multi-scale structure into multi-head self-attention for better speech representation, which restricts the attention area into a large scope at higher layers while restricts the attention area into a small scope at lower layers. Experiments on Librispeech dataset indicate the effectiveness of our proposed method. Compared with HuBERT, PMS-SSL achieves 13.7% / 12.7% relative WER reduction on test other evaluation subsets respectively when fine-tuned on 10hours / 100hours subsets." @default.
- W4310923040 created "2022-12-21" @default.
- W4310923040 creator A5004955508 @default.
- W4310923040 creator A5009590736 @default.
- W4310923040 creator A5011133553 @default.
- W4310923040 creator A5042426677 @default.
- W4310923040 creator A5065761465 @default.
- W4310923040 creator A5077319251 @default.
- W4310923040 date "2022-12-07" @default.
- W4310923040 modified "2023-10-16" @default.
- W4310923040 title "Progressive Multi-Scale Self-Supervised Learning for Speech Recognition" @default.
- W4310923040 doi "https://doi.org/10.48550/arxiv.2212.03480" @default.
- W4310923040 hasPublicationYear "2022" @default.
- W4310923040 type Work @default.
- W4310923040 citedByCount "0" @default.
- W4310923040 crossrefType "posted-content" @default.
- W4310923040 hasAuthorship W4310923040A5004955508 @default.
- W4310923040 hasAuthorship W4310923040A5009590736 @default.
- W4310923040 hasAuthorship W4310923040A5011133553 @default.
- W4310923040 hasAuthorship W4310923040A5042426677 @default.
- W4310923040 hasAuthorship W4310923040A5065761465 @default.
- W4310923040 hasAuthorship W4310923040A5077319251 @default.
- W4310923040 hasBestOaLocation W43109230401 @default.
- W4310923040 hasConcept C111335779 @default.
- W4310923040 hasConcept C121332964 @default.
- W4310923040 hasConcept C153180895 @default.
- W4310923040 hasConcept C154945302 @default.
- W4310923040 hasConcept C17744445 @default.
- W4310923040 hasConcept C178790620 @default.
- W4310923040 hasConcept C185592680 @default.
- W4310923040 hasConcept C199360897 @default.
- W4310923040 hasConcept C199539241 @default.
- W4310923040 hasConcept C2524010 @default.
- W4310923040 hasConcept C2776359362 @default.
- W4310923040 hasConcept C2778012447 @default.
- W4310923040 hasConcept C2778755073 @default.
- W4310923040 hasConcept C2779227376 @default.
- W4310923040 hasConcept C28490314 @default.
- W4310923040 hasConcept C33923547 @default.
- W4310923040 hasConcept C41008148 @default.
- W4310923040 hasConcept C51632099 @default.
- W4310923040 hasConcept C62520636 @default.
- W4310923040 hasConcept C94625758 @default.
- W4310923040 hasConceptScore W4310923040C111335779 @default.
- W4310923040 hasConceptScore W4310923040C121332964 @default.
- W4310923040 hasConceptScore W4310923040C153180895 @default.
- W4310923040 hasConceptScore W4310923040C154945302 @default.
- W4310923040 hasConceptScore W4310923040C17744445 @default.
- W4310923040 hasConceptScore W4310923040C178790620 @default.
- W4310923040 hasConceptScore W4310923040C185592680 @default.
- W4310923040 hasConceptScore W4310923040C199360897 @default.
- W4310923040 hasConceptScore W4310923040C199539241 @default.
- W4310923040 hasConceptScore W4310923040C2524010 @default.
- W4310923040 hasConceptScore W4310923040C2776359362 @default.
- W4310923040 hasConceptScore W4310923040C2778012447 @default.
- W4310923040 hasConceptScore W4310923040C2778755073 @default.
- W4310923040 hasConceptScore W4310923040C2779227376 @default.
- W4310923040 hasConceptScore W4310923040C28490314 @default.
- W4310923040 hasConceptScore W4310923040C33923547 @default.
- W4310923040 hasConceptScore W4310923040C41008148 @default.
- W4310923040 hasConceptScore W4310923040C51632099 @default.
- W4310923040 hasConceptScore W4310923040C62520636 @default.
- W4310923040 hasConceptScore W4310923040C94625758 @default.
- W4310923040 hasLocation W43109230401 @default.
- W4310923040 hasLocation W43109230402 @default.
- W4310923040 hasOpenAccess W4310923040 @default.
- W4310923040 hasPrimaryLocation W43109230401 @default.
- W4310923040 hasRelatedWork W1509972767 @default.
- W4310923040 hasRelatedWork W178231042 @default.
- W4310923040 hasRelatedWork W1932132538 @default.
- W4310923040 hasRelatedWork W2357832196 @default.
- W4310923040 hasRelatedWork W2360028903 @default.
- W4310923040 hasRelatedWork W2362452928 @default.
- W4310923040 hasRelatedWork W2366083136 @default.
- W4310923040 hasRelatedWork W2387622493 @default.
- W4310923040 hasRelatedWork W4200375594 @default.
- W4310923040 hasRelatedWork W4280543773 @default.
- W4310923040 isParatext "false" @default.
- W4310923040 isRetracted "false" @default.
- W4310923040 workType "article" @default.