• Aucun résultat trouvé

End-to-End Automatic Speech Translation of Audiobooks

N/A
N/A
Protected

Academic year: 2021

Partager "End-to-End Automatic Speech Translation of Audiobooks"

Copied!
6
0
0

Texte intégral

Loading

Figure

Table 1: Size of the Augmented LibriSpeech and BTEC corpora, with the average frame, character and word counts (subword count for LibriSpeech) per segment

Références

Documents relatifs

“[m]ore so than any other American author of the twentieth century, Pynchon documents in fiction the tectonic movements of American history, to the extent that

The methodology consists in introducing semantic information by using a class-based statistical language model for which classes directly correspond to IF entries.. With

We apply the average age-specific twinning rate to the observed number of births to predict the overall twinning rates by country by year if only the age distribution of

For a small cloud server cluster (for example, L < 50), increasing the number of user will increase rapidly the response time of request. But for a large cloud server cluster

Table 1: Estimated CO 2 from training a single French (Com- monVoice) or English (LibriSpeech) state-of-the-art end-to-end speech recognizer on Nvidia Tesla V100 GPUs2. Models CO

• IWSLT 2020 simultaneous translation track with a cascade of an ASR system trained using Kaldi (Povey et al., 2011) and an online MT system with wait-k policies (Dalvi et al.,

6. This paper proposes to integrate multiple acous- tic feature views with quaternion hyper complex numbers, and to process these features with a convolutional neural network

In summary, our contributions are: (1) we propose an in- depth study on the impact of self-supervised pre-training for AST, (2) we show that fine-tuning pre-trained representations