Speech:Exps 0288 014

Description
Authors: Team Cap

Date: 4/17/16

Purpose: FINAL NUM DENSITIES has been set to 62 on a 30 hour corpus. This was yet another attempt to see how the WER was affected by upping the density.

Details:
 * Train configuration
 * Corpus: Switchboard
 * Length: 30hr
 * $CFG_FINAL_NUM_DENSITIES = 64 (density)
 * $CFG_N_TIED_STATES = 4000(senones)


 * Timeline
 * Train start: 6:29 PM on 4/17/16
 * Train end: 12:16 AM  on 4/18/16 (Total time: 5:47) ( errors in 014.html: 128)
 * Decode start: 09:15 AM on 4/18/16
 * Decode end: Unknown on 4/18/16 (Total time: Unknown)


 * Decode configuration
 * Decoding on: /mnt/main/corpus/switchboard/30hr/test/trans/dev.trans (~5 hours)
 * Decoding at: 4000 senones to match the senone count in the train configuration

Results: WER: 53.6% Unseen