Matches in SemOpenAlex for { <https://semopenalex.org/work/W2787812568> ?p ?o ?g. }
Showing items 1 to 83 of
83
with 100 items per page.
- W2787812568 abstract "Use of both manually and automatically labelled data for model training is referred to as semi-supervised training. While semi-supervised acoustic model training has been well-explored in the context of hidden Markov Gaussian mixture models (HMM-GMMs), the re-emergence of deep neural network (DNN) acoustic models has given rise to some novel approaches to semi-supervised DNN training. This paper investigates several different strategies for semi-supervised DNN training, including the so-called ‘shared hidden layer’ approach and the ‘knowledge distillation’ (or student-teacher) approach. Particular attention is paid to the differing behaviour of semi-supervised DNN training methods during the cross-entropy and sequence training phases of model building. Experimental results on our internal study dataset provide evidence that in a low-resource scenario the most effective semi-supervised training strategy is ‘naive CE’ (treating manually transcribed and automatically transcribed data identically during the cross entropy phase of training) followed by use of a shared hidden layer technique during sequence training." @default.
- W2787812568 created "2018-02-23" @default.
- W2787812568 creator A5020335543 @default.
- W2787812568 creator A5072972387 @default.
- W2787812568 creator A5090233728 @default.
- W2787812568 date "2017-12-01" @default.
- W2787812568 modified "2023-09-25" @default.
- W2787812568 title "Semi-supervised training strategies for deep neural networks" @default.
- W2787812568 cites W1596641231 @default.
- W2787812568 cites W1975113979 @default.
- W2787812568 cites W2025198378 @default.
- W2787812568 cites W2039151035 @default.
- W2787812568 cites W2102865215 @default.
- W2787812568 cites W2124558353 @default.
- W2787812568 cites W2127499922 @default.
- W2787812568 cites W2143577772 @default.
- W2787812568 cites W2594610113 @default.
- W2787812568 cites W3149335959 @default.
- W2787812568 doi "https://doi.org/10.1109/asru.2017.8268919" @default.
- W2787812568 hasPublicationYear "2017" @default.
- W2787812568 type Work @default.
- W2787812568 sameAs 2787812568 @default.
- W2787812568 citedByCount "1" @default.
- W2787812568 countsByYear W27878125682020 @default.
- W2787812568 crossrefType "proceedings-article" @default.
- W2787812568 hasAuthorship W2787812568A5020335543 @default.
- W2787812568 hasAuthorship W2787812568A5072972387 @default.
- W2787812568 hasAuthorship W2787812568A5090233728 @default.
- W2787812568 hasConcept C106301342 @default.
- W2787812568 hasConcept C119857082 @default.
- W2787812568 hasConcept C121332964 @default.
- W2787812568 hasConcept C136389625 @default.
- W2787812568 hasConcept C153180895 @default.
- W2787812568 hasConcept C153294291 @default.
- W2787812568 hasConcept C154945302 @default.
- W2787812568 hasConcept C23224414 @default.
- W2787812568 hasConcept C2777211547 @default.
- W2787812568 hasConcept C28490314 @default.
- W2787812568 hasConcept C41008148 @default.
- W2787812568 hasConcept C50644808 @default.
- W2787812568 hasConcept C61224824 @default.
- W2787812568 hasConcept C62520636 @default.
- W2787812568 hasConceptScore W2787812568C106301342 @default.
- W2787812568 hasConceptScore W2787812568C119857082 @default.
- W2787812568 hasConceptScore W2787812568C121332964 @default.
- W2787812568 hasConceptScore W2787812568C136389625 @default.
- W2787812568 hasConceptScore W2787812568C153180895 @default.
- W2787812568 hasConceptScore W2787812568C153294291 @default.
- W2787812568 hasConceptScore W2787812568C154945302 @default.
- W2787812568 hasConceptScore W2787812568C23224414 @default.
- W2787812568 hasConceptScore W2787812568C2777211547 @default.
- W2787812568 hasConceptScore W2787812568C28490314 @default.
- W2787812568 hasConceptScore W2787812568C41008148 @default.
- W2787812568 hasConceptScore W2787812568C50644808 @default.
- W2787812568 hasConceptScore W2787812568C61224824 @default.
- W2787812568 hasConceptScore W2787812568C62520636 @default.
- W2787812568 hasLocation W27878125681 @default.
- W2787812568 hasOpenAccess W2787812568 @default.
- W2787812568 hasPrimaryLocation W27878125681 @default.
- W2787812568 hasRelatedWork W2412035906 @default.
- W2787812568 hasRelatedWork W2515319207 @default.
- W2787812568 hasRelatedWork W2785554650 @default.
- W2787812568 hasRelatedWork W2891725580 @default.
- W2787812568 hasRelatedWork W2895771689 @default.
- W2787812568 hasRelatedWork W2904440514 @default.
- W2787812568 hasRelatedWork W2944030659 @default.
- W2787812568 hasRelatedWork W2946856970 @default.
- W2787812568 hasRelatedWork W2963054573 @default.
- W2787812568 hasRelatedWork W3023534255 @default.
- W2787812568 hasRelatedWork W3081976092 @default.
- W2787812568 hasRelatedWork W3113757710 @default.
- W2787812568 hasRelatedWork W3138768133 @default.
- W2787812568 hasRelatedWork W3148060700 @default.
- W2787812568 hasRelatedWork W3149506872 @default.
- W2787812568 hasRelatedWork W3152160377 @default.
- W2787812568 hasRelatedWork W3169165652 @default.
- W2787812568 hasRelatedWork W3170205569 @default.
- W2787812568 hasRelatedWork W3177540946 @default.
- W2787812568 hasRelatedWork W640713170 @default.
- W2787812568 isParatext "false" @default.
- W2787812568 isRetracted "false" @default.
- W2787812568 magId "2787812568" @default.
- W2787812568 workType "article" @default.