FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4319862418> ?p ?o ?g. }

• W4319862418 abstract "Text-only adaptation of a transducer model remains challenging for end-to-end speech recognition since the transducer has no clearly separated acoustic model (AM), language model (LM) or blank model. In this work, we propose a modular hybrid autoregressive transducer (MHAT) that has structurally separated label and blank decoders to predict label and blank distributions, respectively, along with a shared acoustic encoder. The encoder and label decoder outputs are directly projected to AM and internal LM scores and then added to compute label posteriors. We train MHAT with an internal LM loss and a HAT loss to ensure that its internal LM becomes a standalone neural LM that can be effectively adapted to text. Moreover, text adaptation of MHAT fosters a much better LM fusion than internal LM subtraction-based methods. On Google's large-scale production data, a multi-domain MHAT adapted with 100B sentences achieves relative WER reductions of up to 12.4% without LM fusion and 21.5% with LM fusion from 400K-hour trained HAT." @default.
• W4319862418 created "2023-02-11" @default.
• W4319862418 creator A5015927589 @default.
• W4319862418 creator A5025935520 @default.
• W4319862418 creator A5027399854 @default.
• W4319862418 creator A5032640894 @default.
• W4319862418 creator A5032931723 @default.
• W4319862418 creator A5033093547 @default.
• W4319862418 creator A5039108749 @default.
• W4319862418 creator A5040131037 @default.
• W4319862418 creator A5040449061 @default.
• W4319862418 creator A5051392570 @default.
• W4319862418 creator A5071715737 @default.
• W4319862418 creator A5071773009 @default.
• W4319862418 creator A5076026485 @default.
• W4319862418 date "2023-01-09" @default.
• W4319862418 modified "2023-10-16" @default.
• W4319862418 title "Modular Hybrid Autoregressive Transducer" @default.
• W4319862418 cites W1494198834 @default.
• W4319862418 cites W1513820424 @default.
• W4319862418 cites W1558606374 @default.
• W4319862418 cites W1965555277 @default.
• W4319862418 cites W1989549063 @default.
• W4319862418 cites W2010362084 @default.
• W4319862418 cites W2080005694 @default.
• W4319862418 cites W2121879602 @default.
• W4319862418 cites W2127141656 @default.
• W4319862418 cites W2327501763 @default.
• W4319862418 cites W2510867321 @default.
• W4319862418 cites W2515801922 @default.
• W4319862418 cites W2795867901 @default.
• W4319862418 cites W2796339975 @default.
• W4319862418 cites W2911629330 @default.
• W4319862418 cites W2936252403 @default.
• W4319862418 cites W2939164678 @default.
• W4319862418 cites W2972621414 @default.
• W4319862418 cites W2972880214 @default.
• W4319862418 cites W2973122799 @default.
• W4319862418 cites W3008008574 @default.
• W4319862418 cites W3008037978 @default.
• W4319862418 cites W3008181812 @default.
• W4319862418 cites W3015194534 @default.
• W4319862418 cites W3080248383 @default.
• W4319862418 cites W3094667432 @default.
• W4319862418 cites W3097777922 @default.
• W4319862418 cites W3097890746 @default.
• W4319862418 cites W3152221657 @default.
• W4319862418 cites W3163203022 @default.
• W4319862418 cites W3163560333 @default.
• W4319862418 cites W3197661863 @default.
• W4319862418 cites W3197976839 @default.
• W4319862418 cites W3202419788 @default.
• W4319862418 cites W4225307083 @default.
• W4319862418 doi "https://doi.org/10.1109/slt54892.2023.10023194" @default.
• W4319862418 hasPublicationYear "2023" @default.
• W4319862418 type Work @default.
• W4319862418 citedByCount "0" @default.
• W4319862418 crossrefType "proceedings-article" @default.
• W4319862418 hasAuthorship W4319862418A5015927589 @default.
• W4319862418 hasAuthorship W4319862418A5025935520 @default.
• W4319862418 hasAuthorship W4319862418A5027399854 @default.
• W4319862418 hasAuthorship W4319862418A5032640894 @default.
• W4319862418 hasAuthorship W4319862418A5032931723 @default.
• W4319862418 hasAuthorship W4319862418A5033093547 @default.
• W4319862418 hasAuthorship W4319862418A5039108749 @default.
• W4319862418 hasAuthorship W4319862418A5040131037 @default.
• W4319862418 hasAuthorship W4319862418A5040449061 @default.
• W4319862418 hasAuthorship W4319862418A5051392570 @default.
• W4319862418 hasAuthorship W4319862418A5071715737 @default.
• W4319862418 hasAuthorship W4319862418A5071773009 @default.
• W4319862418 hasAuthorship W4319862418A5076026485 @default.
• W4319862418 hasConcept C101468663 @default.
• W4319862418 hasConcept C111919701 @default.
• W4319862418 hasConcept C118505674 @default.
• W4319862418 hasConcept C121332964 @default.
• W4319862418 hasConcept C127413603 @default.
• W4319862418 hasConcept C134306372 @default.
• W4319862418 hasConcept C149782125 @default.
• W4319862418 hasConcept C153180895 @default.
• W4319862418 hasConcept C154945302 @default.
• W4319862418 hasConcept C159877910 @default.
• W4319862418 hasConcept C24890656 @default.
• W4319862418 hasConcept C2778089247 @default.
• W4319862418 hasConcept C28490314 @default.
• W4319862418 hasConcept C33923547 @default.
• W4319862418 hasConcept C36503486 @default.
• W4319862418 hasConcept C41008148 @default.
• W4319862418 hasConcept C56318395 @default.
• W4319862418 hasConcept C78519656 @default.
• W4319862418 hasConceptScore W4319862418C101468663 @default.
• W4319862418 hasConceptScore W4319862418C111919701 @default.
• W4319862418 hasConceptScore W4319862418C118505674 @default.
• W4319862418 hasConceptScore W4319862418C121332964 @default.
• W4319862418 hasConceptScore W4319862418C127413603 @default.
• W4319862418 hasConceptScore W4319862418C134306372 @default.
• W4319862418 hasConceptScore W4319862418C149782125 @default.
• W4319862418 hasConceptScore W4319862418C153180895 @default.
• W4319862418 hasConceptScore W4319862418C154945302 @default.
• W4319862418 hasConceptScore W4319862418C159877910 @default.
• W4319862418 hasConceptScore W4319862418C24890656 @default.

• next