Mengyuan Zhao 2015-12-28

From cslt Wiki
Revision as of 13:26, 27 December 2015 by Zhaomy (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

RNN AM

  • AM training on big dataset --mengyuan
  • Further test nnet3-ctc training on 4000h 8k & 10000h 16k dataset. Tune decoding configures (phone-lm-weight, acwt, blank-scale), but ctc is still about 5 percent worse than standard nnet3.