2 |
Automatic Speech Recognition systems errors for accident-prone sleepiness detection through voice
|
|
|
|
In: EUSIPCO 2021 ; https://hal.archives-ouvertes.fr/hal-03324033 ; EUSIPCO 2021, Aug 2021, Dublin (en ligne), Ireland. ⟨10.23919/EUSIPCO54536.2021.9616299⟩ (2021)
|
|
BASE
|
|
Show details
|
|
3 |
Automatic Speech Recognition systems errors for objective sleepiness detection through voice
|
|
|
|
In: Proceedings Interspeech 2021 ; Interspeech 2021 ; https://hal.archives-ouvertes.fr/hal-03328827 ; Interspeech 2021, Aug 2021, Brno (virtual), Czech Republic. pp.2476-2480, ⟨10.21437/Interspeech.2021-291⟩ (2021)
|
|
BASE
|
|
Show details
|
|
4 |
LeBenchmark: A Reproducible Framework for Assessing Self-Supervised Representation Learning from Speech
|
|
|
|
In: INTERSPEECH 2021: Conference of the International Speech Communication Association ; https://hal.archives-ouvertes.fr/hal-03317730 ; INTERSPEECH 2021: Conference of the International Speech Communication Association, Aug 2021, Brno, Czech Republic (2021)
|
|
BASE
|
|
Show details
|
|
5 |
LeBenchmark: A Reproducible Framework for Assessing Self-Supervised Representation Learning from Speech
|
|
|
|
In: INTERSPEECH 2021: ; INTERSPEECH 2021: Conference of the International Speech Communication Association ; https://hal.archives-ouvertes.fr/hal-03317730 ; INTERSPEECH 2021: Conference of the International Speech Communication Association, Aug 2021, Brno, Czech Republic (2021)
|
|
BASE
|
|
Show details
|
|
6 |
LeBenchmark: A Reproducible Framework for Assessing Self-Supervised Representation Learning from Speech
|
|
|
|
In: INTERSPEECH 2021: ; INTERSPEECH 2021: Conference of the International Speech Communication Association ; https://hal.archives-ouvertes.fr/hal-03317730 ; INTERSPEECH 2021: Conference of the International Speech Communication Association, Aug 2021, Brno, Czech Republic (2021)
|
|
BASE
|
|
Show details
|
|
7 |
Re-synchronization using the Hand Preceding Model for Multi-modal Fusion in Automatic Continuous Cued Speech Recognition
|
|
|
|
In: ISSN: 1520-9210 ; IEEE Transactions on Multimedia ; https://hal.archives-ouvertes.fr/hal-02433830 ; IEEE Transactions on Multimedia, Institute of Electrical and Electronics Engineers, 2021, 23, pp.292-305. ⟨10.1109/TMM.2020.2976493⟩ (2021)
|
|
BASE
|
|
Show details
|
|
9 |
Unsupervised Morphological Segmentation and Part-of-Speech Tagging for Low-Resource Scenarios
|
|
|
|
BASE
|
|
Show details
|
|
10 |
Recognizing lexical units in low-resource language contexts with supervised and unsupervised neural networks
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03429051 ; [Research Report] LACITO (UMR 7107). 2021 (2021)
|
|
BASE
|
|
Show details
|
|
11 |
Speech Normalization and Data Augmentation Techniques Based on Acoustical and Physiological Constraints and Their Applications to Child Speech Recognition
|
|
|
|
BASE
|
|
Show details
|
|
12 |
Large vocabulary automatic speech recognition: from hybrid to end-to-end approaches ; Reconnaissance automatique de la parole à large vocabulaire : des approches hybrides aux approches End-to-End
|
|
|
|
In: https://hal.archives-ouvertes.fr/tel-03269807 ; Son [cs.SD]. Université toulouse 3 Paul Sabatier, 2021. Français (2021)
|
|
BASE
|
|
Show details
|
|
13 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
14 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
15 |
The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03332224 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
16 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
17 |
The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03332224 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
18 |
Enhancing Speech Privacy with Slicing
|
|
|
|
In: https://hal.inria.fr/hal-03369137 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
19 |
Supplementary material to the paper The VoicePrivacy 2020 Challenge: Results and findings
|
|
|
|
In: https://hal.archives-ouvertes.fr/hal-03335126 ; 2021 (2021)
|
|
BASE
|
|
Show details
|
|
20 |
Training RNN Language Models on Uncertain ASR Hypotheses in Limited Data Scenarios
|
|
|
|
In: https://hal.inria.fr/hal-03327306 ; 2021 (2021)
|
|
Abstract:
Training domain-specific automatic speech recognition (ASR) systems requires a suitable amount of data comprising the target domain. In several scenarios, such as early development stages, privacy-critical applications, or under-resourced languages, only a limited amount of in-domain speech data and an even smaller amount of manual text transcriptions, if any, are available. This motivates the study of ASR language models (LMs) learned from a limited amount of in-domain speech data. Early works have attempted training of n-gram LMs from ASR N-best lists and lattices but training and adaptation of recurrent neural network (RNN) LMs from ASR transcripts has not received attention. In this work, we study training and adaptation of RNN LMs using alternate and uncertain ASR hypotheses embedded in ASR confusion networks obtained from target domain speech data.We explore different methods for training the RNN LMs to deal with the uncertain input sequences. The first method extends the cross-entropy objective into a Kullback–Leibler (KL) divergence based training loss, the second method formulates a training loss based on a hidden Markov model (HMM), and the third method performs training on paths sampled from the confusion networks. These methods are applied to limited data setups including telephone and meeting conversation datasets. Performance is evaluated under two settings wherein no manual transcriptions or a small amount of manual transcriptions are available to aid the training. Moreover, a model adaptation setting is also evaluated wherein the RNN LM is pre-trained on an out-of-domain conversational corpus. Overall the sampling method for training RNN LMs on ASR confusion networks performs the best, and results in up to 12% relative reduction in perplexity on the meeting dataset as compared to training on ASR 1-best hypotheses, without any manual transcriptions.
|
|
Keyword:
[INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]; [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG]; automatic speech recognition; confusion networks; language models; recurrent neural networks
|
|
URL: https://hal.inria.fr/hal-03327306/document https://hal.inria.fr/hal-03327306/file/cn2lm_manuscript.pdf https://hal.inria.fr/hal-03327306
|
|
BASE
|
|
Hide details
|
|
|
|