Speech:Exps 0172

From Openitware
Jump to: navigation, search

Description

Author: David Meehan

Date: 2/18/2014

Purpose: Running a five hour train/decode to attempt to replicate the results Colby got in experiment 0175. This will be the acoustic model and language model I will be using for a number of future decodes using test data.

Details:

  • Corpus: first_5hr/train
  • Senone value: 3000
  • Density: 64


Results

                    SYSTEM SUMMARY PERCENTAGES by SPEAKER                      
     ,-----------------------------------------------------------------.
     |                            hyp.trans                            |
     |-----------------------------------------------------------------|
     | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err |
     |=================================================================|
     | Sum/Avg | 4659  68616 | 93.6    3.3    3.1    8.8   15.2   71.5 |
     |=================================================================|
     |  Mean   | 58.2  857.7 | 93.6    3.4    3.0    9.7   16.2   73.5 |
     |  S.D.   | 22.1  330.0 |  2.6    1.4    2.0    5.1    6.1   11.0 |
     | Median  | 55.5  813.0 | 93.9    3.1    2.6    9.2   15.0   75.0 |
     `-----------------------------------------------------------------'

Successful Completion