no code implementations • 2 Jul 2019 • Mortaza, Doulaty, Thomas Hain
The proposed technique for training data selection, significantly outperforms random selection, posterior-based selection as well as using all of the available data.
2k Automatic Speech Recognition +5