Matches in SemOpenAlex for { <https://semopenalex.org/work/W3132012095> ?p ?o ?g. }
- W3132012095 abstract "Spoken language understanding (SLU) systems extract transcriptions, as well as semantics of intent or named entities from speech, and are essential components of voice activated systems. SLU models, which either directly extract semantics from audio or are composed of pipelined automatic speech recognition (ASR) and natural language understanding (NLU) models, are typically trained via differentiable cross-entropy losses, even when the relevant performance metrics of interest are word or semantic error rates. In this work, we propose non-differentiable sequence losses based on SLU metrics as a proxy for semantic error and use the REINFORCE trick to train ASR and SLU models with this loss. We show that custom sequence loss training is the state-of-the-art on open SLU datasets and leads to 6% relative improvement in both ASR and NLU performance metrics on large proprietary datasets. We also demonstrate how the semantic sequence loss training paradigm can be used to update ASR and SLU models without transcripts, using semantic feedback alone." @default.
- W3132012095 created "2021-03-01" @default.
- W3132012095 creator A5003334789 @default.
- W3132012095 creator A5008201430 @default.
- W3132012095 creator A5012153296 @default.
- W3132012095 creator A5031236216 @default.
- W3132012095 creator A5060261873 @default.
- W3132012095 creator A5060979948 @default.
- W3132012095 creator A5076504502 @default.
- W3132012095 date "2021-02-12" @default.
- W3132012095 modified "2023-09-26" @default.
- W3132012095 title "Do as I mean, not as I say: Sequence Loss Training for Spoken Language Understanding" @default.
- W3132012095 cites W1828163288 @default.
- W3132012095 cites W1989996186 @default.
- W3132012095 cites W2007261869 @default.
- W3132012095 cites W2016589492 @default.
- W3132012095 cites W2119717200 @default.
- W3132012095 cites W2127141656 @default.
- W3132012095 cites W2158289097 @default.
- W3132012095 cites W2296283641 @default.
- W3132012095 cites W2327501763 @default.
- W3132012095 cites W2547875792 @default.
- W3132012095 cites W2786839803 @default.
- W3132012095 cites W2787045460 @default.
- W3132012095 cites W2894164357 @default.
- W3132012095 cites W2914417638 @default.
- W3132012095 cites W2917128112 @default.
- W3132012095 cites W2928075308 @default.
- W3132012095 cites W2962715022 @default.
- W3132012095 cites W2962824709 @default.
- W3132012095 cites W2963288440 @default.
- W3132012095 cites W2963747784 @default.
- W3132012095 cites W2968831808 @default.
- W3132012095 cites W3005910077 @default.
- W3132012095 cites W3006901707 @default.
- W3132012095 cites W3017465475 @default.
- W3132012095 cites W3049038774 @default.
- W3132012095 cites W3094979069 @default.
- W3132012095 cites W3095552229 @default.
- W3132012095 cites W3113320605 @default.
- W3132012095 cites W55328212 @default.
- W3132012095 doi "https://doi.org/10.48550/arxiv.2102.06750" @default.
- W3132012095 hasPublicationYear "2021" @default.
- W3132012095 type Work @default.
- W3132012095 sameAs 3132012095 @default.
- W3132012095 citedByCount "0" @default.
- W3132012095 crossrefType "posted-content" @default.
- W3132012095 hasAuthorship W3132012095A5003334789 @default.
- W3132012095 hasAuthorship W3132012095A5008201430 @default.
- W3132012095 hasAuthorship W3132012095A5012153296 @default.
- W3132012095 hasAuthorship W3132012095A5031236216 @default.
- W3132012095 hasAuthorship W3132012095A5060261873 @default.
- W3132012095 hasAuthorship W3132012095A5060979948 @default.
- W3132012095 hasAuthorship W3132012095A5076504502 @default.
- W3132012095 hasBestOaLocation W31320120951 @default.
- W3132012095 hasConcept C119857082 @default.
- W3132012095 hasConcept C138885662 @default.
- W3132012095 hasConcept C154945302 @default.
- W3132012095 hasConcept C162324750 @default.
- W3132012095 hasConcept C184337299 @default.
- W3132012095 hasConcept C187736073 @default.
- W3132012095 hasConcept C195324797 @default.
- W3132012095 hasConcept C199360897 @default.
- W3132012095 hasConcept C204321447 @default.
- W3132012095 hasConcept C2776230583 @default.
- W3132012095 hasConcept C2778112365 @default.
- W3132012095 hasConcept C2779439875 @default.
- W3132012095 hasConcept C2780148112 @default.
- W3132012095 hasConcept C2780451532 @default.
- W3132012095 hasConcept C28490314 @default.
- W3132012095 hasConcept C35639132 @default.
- W3132012095 hasConcept C41008148 @default.
- W3132012095 hasConcept C41895202 @default.
- W3132012095 hasConcept C54355233 @default.
- W3132012095 hasConcept C86803240 @default.
- W3132012095 hasConcept C90805587 @default.
- W3132012095 hasConceptScore W3132012095C119857082 @default.
- W3132012095 hasConceptScore W3132012095C138885662 @default.
- W3132012095 hasConceptScore W3132012095C154945302 @default.
- W3132012095 hasConceptScore W3132012095C162324750 @default.
- W3132012095 hasConceptScore W3132012095C184337299 @default.
- W3132012095 hasConceptScore W3132012095C187736073 @default.
- W3132012095 hasConceptScore W3132012095C195324797 @default.
- W3132012095 hasConceptScore W3132012095C199360897 @default.
- W3132012095 hasConceptScore W3132012095C204321447 @default.
- W3132012095 hasConceptScore W3132012095C2776230583 @default.
- W3132012095 hasConceptScore W3132012095C2778112365 @default.
- W3132012095 hasConceptScore W3132012095C2779439875 @default.
- W3132012095 hasConceptScore W3132012095C2780148112 @default.
- W3132012095 hasConceptScore W3132012095C2780451532 @default.
- W3132012095 hasConceptScore W3132012095C28490314 @default.
- W3132012095 hasConceptScore W3132012095C35639132 @default.
- W3132012095 hasConceptScore W3132012095C41008148 @default.
- W3132012095 hasConceptScore W3132012095C41895202 @default.
- W3132012095 hasConceptScore W3132012095C54355233 @default.
- W3132012095 hasConceptScore W3132012095C86803240 @default.
- W3132012095 hasConceptScore W3132012095C90805587 @default.
- W3132012095 hasLocation W31320120951 @default.
- W3132012095 hasOpenAccess W3132012095 @default.
- W3132012095 hasPrimaryLocation W31320120951 @default.