Speech:Exps 0109

From Openitware
Jump to: navigation, search

Modified Senone Value (3000): Test on Train 0106


Description

Author: Eric Beikman

Date: 6/20/2013

Purpose: To run a test on train for Experiment 0106.

Details: For this Experiment, we utilized the 30 minute last_5hr/test corpus to test train 0106. In Experiment 0106, we increased the amount of senones to 3000 to determine the effect on the resulting model.


To enable us to compare to our previous "Best" scoring experiment for this corpus (Experiment 0090), this experiment utilizes the language model, dictionaries, and phone list from that particular experiment. As such, this also allows us to compare the generated score with Experiment 0104, which had also had a modified senone value. Results

The following score was created during this experiment:

       ,-----------------------------------------------------------------.
      |                            hyp.trans                            |
      |-----------------------------------------------------------------|
      | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err |
      |---------+-------------+-----------------------------------------|
      | Sum/Avg |  437   6474 | 87.9    8.5    3.6   14.3   26.4   95.0 |
      |=================================================================|
      |  Mean   | 36.4  539.5 | 87.7    8.7    3.6   15.5   27.7   95.1 |
      |  S.D.   |  8.3  143.2 |  3.4    2.7    1.3    6.5    7.7    4.7 |
      | Median  | 32.5  546.5 | 87.6    9.2    3.4   15.0   27.9   95.4 |
      `-----------------------------------------------------------------'

This experiment sets a new record for the word error rate for this corpus. By slightly increasing the senone value (by 1000), we were able to get about a 1 point improvement in our word error rate; however, this is less than increasing the senone value from 2000 to 2500.

Please see the results of [[Speech:Exps_0110|Experiment 0110] for a more in-depth analysis.