Deep Speech 2: End-to-End Speech Recognition in English and Mandarin
At a glance
Neural networks transcribed English and Mandarin speech, matching human transcription on some benchmarks.
In one paragraph
Extended end-to-end speech recognition to English and Mandarin, learning to transcribe audio with neural networks. Faster training enabled larger experiments, while efficient deployment supported online use. Matched human transcription on some benchmarks. Published at ICML 2016; first released as a preprint in 2015.
Authors 69
Organizations
Credited for this work at publication.