Matches in SemOpenAlex for { <https://semopenalex.org/work/W2912691643> ?p ?o ?g. }
Showing items 1 to 81 of
81
with 100 items per page.
- W2912691643 abstract "Traditional LSTM model and its variants normally work in a frame-by-frame and layer-by-layer fashion, which deals with the temporal modeling and target classification problems at the same time. In this paper, we extend our recently proposed layer trajectory LSTM (ltLSTM) and present a generalized framework, which is equipped with a depth processing block that scans the hidden states of each time-LSTM layer, and uses the summarized layer trajectory information for final senone classification. We explore different modeling units used in the depth processing block to have a good tradeoff between accuracy and runtime cost. Furthermore, we integrate an attention module into this framework to explore wide context information, which is especially beneficial for uni-directional LSTMs. Trained with 30 thousand hours of EN-US Microsoft internal data and cross entropy criterion, the proposed generalized ltLSTM performed significantly better than the standard multi-layer time-LSTM, with up to 12.8% relative word error rate (WER) reduction across different tasks. With attention modeling, the relative WER reduction can be up to 17.9%. We observed similar gain when the models were trained with sequence discriminative training criterion." @default.
- W2912691643 created "2019-02-21" @default.
- W2912691643 creator A5022890229 @default.
- W2912691643 creator A5029670581 @default.
- W2912691643 creator A5030625154 @default.
- W2912691643 creator A5077401426 @default.
- W2912691643 date "2018-12-01" @default.
- W2912691643 modified "2023-09-23" @default.
- W2912691643 title "Exploring Layer Trajectory LSTM with Depth Processing Units and Attention" @default.
- W2912691643 cites W1600744878 @default.
- W2912691643 cites W1984541135 @default.
- W2912691643 cites W1987238397 @default.
- W2912691643 cites W2061305730 @default.
- W2912691643 cites W2064675550 @default.
- W2912691643 cites W2076794394 @default.
- W2912691643 cites W2089917322 @default.
- W2912691643 cites W2119615570 @default.
- W2912691643 cites W2131342762 @default.
- W2912691643 cites W2143612262 @default.
- W2912691643 cites W2160815625 @default.
- W2912691643 cites W2296748324 @default.
- W2912691643 cites W2396144419 @default.
- W2912691643 cites W2396384435 @default.
- W2912691643 cites W2475988411 @default.
- W2912691643 cites W2513938599 @default.
- W2912691643 cites W2586630418 @default.
- W2912691643 cites W2734724284 @default.
- W2912691643 cites W2735006420 @default.
- W2912691643 cites W2747006787 @default.
- W2912691643 cites W2787420813 @default.
- W2912691643 cites W2962962542 @default.
- W2912691643 cites W2963326356 @default.
- W2912691643 cites W2963523217 @default.
- W2912691643 cites W2964084166 @default.
- W2912691643 doi "https://doi.org/10.1109/slt.2018.8639637" @default.
- W2912691643 hasPublicationYear "2018" @default.
- W2912691643 type Work @default.
- W2912691643 sameAs 2912691643 @default.
- W2912691643 citedByCount "6" @default.
- W2912691643 countsByYear W29126916432018 @default.
- W2912691643 countsByYear W29126916432019 @default.
- W2912691643 countsByYear W29126916432021 @default.
- W2912691643 countsByYear W29126916432022 @default.
- W2912691643 crossrefType "proceedings-article" @default.
- W2912691643 hasAuthorship W2912691643A5022890229 @default.
- W2912691643 hasAuthorship W2912691643A5029670581 @default.
- W2912691643 hasAuthorship W2912691643A5030625154 @default.
- W2912691643 hasAuthorship W2912691643A5077401426 @default.
- W2912691643 hasConcept C121332964 @default.
- W2912691643 hasConcept C1276947 @default.
- W2912691643 hasConcept C13662910 @default.
- W2912691643 hasConcept C154945302 @default.
- W2912691643 hasConcept C159985019 @default.
- W2912691643 hasConcept C192562407 @default.
- W2912691643 hasConcept C2779227376 @default.
- W2912691643 hasConcept C41008148 @default.
- W2912691643 hasConceptScore W2912691643C121332964 @default.
- W2912691643 hasConceptScore W2912691643C1276947 @default.
- W2912691643 hasConceptScore W2912691643C13662910 @default.
- W2912691643 hasConceptScore W2912691643C154945302 @default.
- W2912691643 hasConceptScore W2912691643C159985019 @default.
- W2912691643 hasConceptScore W2912691643C192562407 @default.
- W2912691643 hasConceptScore W2912691643C2779227376 @default.
- W2912691643 hasConceptScore W2912691643C41008148 @default.
- W2912691643 hasLocation W29126916431 @default.
- W2912691643 hasOpenAccess W2912691643 @default.
- W2912691643 hasPrimaryLocation W29126916431 @default.
- W2912691643 hasRelatedWork W1562288862 @default.
- W2912691643 hasRelatedWork W1571484390 @default.
- W2912691643 hasRelatedWork W1978753601 @default.
- W2912691643 hasRelatedWork W2000407620 @default.
- W2912691643 hasRelatedWork W2007815619 @default.
- W2912691643 hasRelatedWork W2095989223 @default.
- W2912691643 hasRelatedWork W2164159849 @default.
- W2912691643 hasRelatedWork W2907770832 @default.
- W2912691643 hasRelatedWork W2913211185 @default.
- W2912691643 hasRelatedWork W4246474087 @default.
- W2912691643 isParatext "false" @default.
- W2912691643 isRetracted "false" @default.
- W2912691643 magId "2912691643" @default.
- W2912691643 workType "article" @default.