Speech:Exps 0026

From Openitware
Jump to: navigation, search

Mini Eval Decode. Part 1


Description

Author: Eric Beikman

Date: (Started) 3/18/2013

Purpose: To verify the results in Experiment 0024 and 0025 in which models that were created with redundant transcript entries are slightly more accurate than models created with only unique transcript entries.

Details: To begin this experiment, the experiment directory was set up like a typical training session. We will be decoding using the mini/eval corpus subset. Since we are using the acoustic model of Experiment 0024, an Train was not completed for this experiment. Due to its necessity for the decoding process, a Language model for the mini/eval corpus subset was created for this experiment.


Results The following score was created:

                     SYSTEM SUMMARY PERCENTAGES by SPEAKER

      ,-----------------------------------------------------------------.
      |                            hyp.trans                            |
      |-----------------------------------------------------------------|
      | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err |
      |=================================================================|
      | Sum/Avg |  518  11051 | 44.4   45.6    9.9   10.6   66.2  100.0 |
      |=================================================================|
      |  Mean   |  2.7   58.2 | 45.5   46.3    8.2   15.7   70.1  100.0 |
      |  S.D.   |  1.7   43.0 | 13.2   13.0    7.2   19.3   19.7    0.0 |
      | Median  |  2.0   52.5 | 44.0   47.3    7.2   10.3   68.1  100.0 |
      `-----------------------------------------------------------------'

This experiment and a summary will be continued in Speech:Exps_0053