ICASSP
2011
IEEE
12 years 8 months ago
2011
IEEE
Traditionally, the use of untranscribed speech has been restricted to unsupervised or semi-supervised training of acoustic models. Comparison of recognizers has required labeled d...