Speech:Exps 0090

From Openitware
Jump to: navigation, search

Last_5hr Test on Train (0089) using modified transcript for LM creation.


Description

Author:Team B&C

Date: 4/22/2013

Purpose: This experiment was designed to test the newly created acoustic model from 0089 and to test using a Language model derived from genTrans5.pl.

Details: The previous versions of the genTrans script merely removed markers which indicated transcribers' notes, leaving whatever was in those markers intact. This resulted in inaccurate models as the trainer was attempting to align and train phones listed in the transcript but did not exist in the audio file.

As the existing processes for creating the transcript used for the Language model creation suffered from the same issues described above, the language models it created were likely inaccurate as well as the grammar structures it was analyzing was not representative of normal English.

The transcript used in this experiment was instead derived from the genTrans5-generated transcript from experiment 0089. Sed was then used to prepare the transcript in a manner similar to the existing ParseTranscript.perl script to avoid encountering issues with the script. The script used was as follows:

cat ../../0089/LM/trans_parsed | sed 's/<S>//g' | sed 's/<\/S>//g' | sed 's/()//g' > trans_parsed

The remainder of the Language model creation, decode, and scoring steps were not deviated from.

Results The experiment was successful in that the all steps were completed. The Language model was created successfully using the modified transcript. The following score was produced:



                     SYSTEM SUMMARY PERCENTAGES by SPEAKER

      ,-----------------------------------------------------------------.
      |                              0090                               |
      |-----------------------------------------------------------------|
      | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err |
      |=================================================================|
      | Sum/Avg |  437   6474 | 82.4   13.2    4.4   14.3   31.9   96.1 |
      |=================================================================|
      |  Mean   | 36.4  539.5 | 82.6   13.1    4.3   15.4   32.8   96.3 |
      |  S.D.   |  8.3  143.2 |  4.4    3.4    1.7    6.0    6.9    3.8 |
      | Median  | 32.5  546.5 | 83.4   13.7    4.1   15.4   34.0   97.3 |
      `-----------------------------------------------------------------'

This experiment is continued in Experiment 0091 where we perform a similar decode and score with a Language model created using the existing processes.

The results as compared to the first test decode and score of the acoustic model created using the Last_5hr train corpus are better by a single point.