Matches in SemOpenAlex for { <https://semopenalex.org/work/W4319862463> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W4319862463 abstract "Differential privacy (DP) is one data protection avenue to safeguard user information used for training deep models by imposing noisy distortion on privacy data. Such a noise perturbation often results in a severe performance degradation in automatic speech recognition (ASR) in order to meet a privacy budget ε. Private aggregation of teacher ensemble (PATE) utilizes ensemble probabilities to improve ASR accuracy when dealing with the noise effects controlled by small values of ε. We extend PATE learning to work with dynamic patterns, namely speech utterances, and perform a first experimental demonstration that it prevents acoustic data leakage in ASR training. We evaluate three end-to-end deep models, including LAS, hybrid CTC/attention, and RNN transducer, on the open-source LibriSpeech and TIMIT corpora. PATE learning-enhanced ASR models outperform the benchmark DP-SGD mechanisms, especially under strict DP budgets, giving relative word error rate reductions between 26.2% and 27.5% for an RNN transducer model evaluated with LibriSpeech. We also introduce a DP-preserving ASR solution for pretraining on public speech corpora." @default.
- W4319862463 created "2023-02-11" @default.
- W4319862463 creator A5012653576 @default.
- W4319862463 creator A5020376803 @default.
- W4319862463 creator A5060979948 @default.
- W4319862463 creator A5066868860 @default.
- W4319862463 creator A5079659476 @default.
- W4319862463 date "2023-01-09" @default.
- W4319862463 modified "2023-09-29" @default.
- W4319862463 title "An Experimental Study on Private Aggregation of Teacher Ensemble Learning for End-to-End Speech Recognition" @default.
- W4319862463 cites W1494198834 @default.
- W4319862463 cites W1635512741 @default.
- W4319862463 cites W2048673468 @default.
- W4319862463 cites W2051267297 @default.
- W4319862463 cites W2061533153 @default.
- W4319862463 cites W2084514013 @default.
- W4319862463 cites W2127141656 @default.
- W4319862463 cites W2128970689 @default.
- W4319862463 cites W2148154194 @default.
- W4319862463 cites W2167372639 @default.
- W4319862463 cites W2293829681 @default.
- W4319862463 cites W2327501763 @default.
- W4319862463 cites W2473418344 @default.
- W4319862463 cites W2508418541 @default.
- W4319862463 cites W2526425061 @default.
- W4319862463 cites W2594311007 @default.
- W4319862463 cites W2690039764 @default.
- W4319862463 cites W2806521799 @default.
- W4319862463 cites W2896422817 @default.
- W4319862463 cites W2962826786 @default.
- W4319862463 cites W3025567392 @default.
- W4319862463 cites W3161932608 @default.
- W4319862463 cites W3162072286 @default.
- W4319862463 cites W3173081491 @default.
- W4319862463 cites W4210970267 @default.
- W4319862463 cites W4225262970 @default.
- W4319862463 cites W4226316089 @default.
- W4319862463 cites W4226418308 @default.
- W4319862463 cites W3198781161 @default.
- W4319862463 doi "https://doi.org/10.1109/slt54892.2023.10023326" @default.
- W4319862463 hasPublicationYear "2023" @default.
- W4319862463 type Work @default.
- W4319862463 citedByCount "0" @default.
- W4319862463 crossrefType "proceedings-article" @default.
- W4319862463 hasAuthorship W4319862463A5012653576 @default.
- W4319862463 hasAuthorship W4319862463A5020376803 @default.
- W4319862463 hasAuthorship W4319862463A5060979948 @default.
- W4319862463 hasAuthorship W4319862463A5066868860 @default.
- W4319862463 hasAuthorship W4319862463A5079659476 @default.
- W4319862463 hasConcept C115961682 @default.
- W4319862463 hasConcept C124101348 @default.
- W4319862463 hasConcept C13280743 @default.
- W4319862463 hasConcept C137293760 @default.
- W4319862463 hasConcept C154945302 @default.
- W4319862463 hasConcept C185798385 @default.
- W4319862463 hasConcept C205649164 @default.
- W4319862463 hasConcept C23130292 @default.
- W4319862463 hasConcept C23224414 @default.
- W4319862463 hasConcept C2778724510 @default.
- W4319862463 hasConcept C28490314 @default.
- W4319862463 hasConcept C40969351 @default.
- W4319862463 hasConcept C41008148 @default.
- W4319862463 hasConcept C74296488 @default.
- W4319862463 hasConcept C99498987 @default.
- W4319862463 hasConceptScore W4319862463C115961682 @default.
- W4319862463 hasConceptScore W4319862463C124101348 @default.
- W4319862463 hasConceptScore W4319862463C13280743 @default.
- W4319862463 hasConceptScore W4319862463C137293760 @default.
- W4319862463 hasConceptScore W4319862463C154945302 @default.
- W4319862463 hasConceptScore W4319862463C185798385 @default.
- W4319862463 hasConceptScore W4319862463C205649164 @default.
- W4319862463 hasConceptScore W4319862463C23130292 @default.
- W4319862463 hasConceptScore W4319862463C23224414 @default.
- W4319862463 hasConceptScore W4319862463C2778724510 @default.
- W4319862463 hasConceptScore W4319862463C28490314 @default.
- W4319862463 hasConceptScore W4319862463C40969351 @default.
- W4319862463 hasConceptScore W4319862463C41008148 @default.
- W4319862463 hasConceptScore W4319862463C74296488 @default.
- W4319862463 hasConceptScore W4319862463C99498987 @default.
- W4319862463 hasLocation W43198624631 @default.
- W4319862463 hasOpenAccess W4319862463 @default.
- W4319862463 hasPrimaryLocation W43198624631 @default.
- W4319862463 hasRelatedWork W1006721676 @default.
- W4319862463 hasRelatedWork W1586532344 @default.
- W4319862463 hasRelatedWork W2050397613 @default.
- W4319862463 hasRelatedWork W2106051978 @default.
- W4319862463 hasRelatedWork W2124052683 @default.
- W4319862463 hasRelatedWork W2996122240 @default.
- W4319862463 hasRelatedWork W3011988934 @default.
- W4319862463 hasRelatedWork W3046672974 @default.
- W4319862463 hasRelatedWork W3079443720 @default.
- W4319862463 hasRelatedWork W3175483649 @default.
- W4319862463 isParatext "false" @default.
- W4319862463 isRetracted "false" @default.
- W4319862463 workType "article" @default.