Example: bankruptcy
arXiv:2006.13979v2 [cs.CL] 15 Dec 2020
(§ 2.1), training (§ 2.2) and adaptations for cross-lingual training. 2.1 ARCHITECTURE We follow the design choices described in Baevski et al. (2020c). The model contains a convolutional feature encoder f: X7!Zto map raw audio Xto latent speech representations z 1;:::;z Twhich are fed to a Transformer network g: Z7!Cto output context ...
Tags:
Information
Domain:
Source:
Link to this page: