Speech:Exps 0075

From Openitware
Jump to: navigation, search

Last_5hr LM creation & Test on Train(0074)


Description

Author: Spring 2013: Groups B&C

Date: 4/8 - 4/9/13

Purpose: Create an Language model on a 5 hour set of new corpus data and test decode and score on the acoustic model created in experiment 0075.

Details: The Spring 2013 Data group has found a complete set of transcripts for the Switchboard corpus. This set of transcripts brings our total amount of available audio data to about 308 hours. Our goal for this experiment is to take the last 5 hours of this 308 hours worth of data, and create an language model from it.

After, we will take a 30 minute subset, starting from the beginning of this 5 hour corpus, and and utilize it to run a test on the Language model created as a part of this experiment and the acoustic model created in [Speech:Exps_0074| Experiment 0074].


Results This experiment was a group effort, with various group members completing steps in the process. No issues were found out of the ordinary for most steps.

The experiment dictionary created for Experiment 0074 was used for this experiment.

Like experiment 0074, the words were "<A_ASIDE>" and "<B_ASIDE>" were in the transcript, we believe that these are translators notes which somehow got left in. We edited these out before proceeding with the experiment.

The following score was created:

      ,-----------------------------------------------------------------.
      |                            hyp.trans                            |
      |-----------------------------------------------------------------|
      | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err |
      |---------+-------------+-----------------------------------------|
      | Sum/Avg |  437   6569 | 79.4   15.6    5.0   12.5   33.1   96.1 |
      |=================================================================|
      |  Mean   | 36.4  547.4 | 79.1   16.0    4.9   13.4   34.3   96.0 |
      |  S.D.   |  8.3  142.1 |  4.7    4.0    2.0    5.1    7.6    4.3 |
      | Median  | 32.5  557.0 | 78.9   17.0    4.9   13.2   36.9   96.9 |
      `-----------------------------------------------------------------'