2019 DS/ML digest 12

2019 DS/ML digest 12

Posted by snakers41 on July 3, 2019

STT

  • Almost(?) unsupervised STT on 20 minutes of speech - 3 days of training on large node for 1 voice;
  • A very cool paper in STT - essentially they try to build ResNet like models instead of plain wav2letter like CNNs;


NLP


Posts / articles:


Python cool articles:


Libraries / repos:

  • Dowloading tons of images from Google - you can use this or you can just use seo-utils.ru + download images from links later;


Internet / news


Datasets


The problems with academic world