Speech:Exps 0314 004

From Openitware
Jump to: navigation, search

Description

Author: Monica (UserID: mlp1025)

Date: 3-5-2019

Purpose: Re-runing 0305/012 using current genTrans.pl Blah blah blah

Details: This experiment is intended to be a baseline experiment for a 5 hour training corpus using 5 hours of test-on-train corpus (i.e. the same 5 hours we trained on). It is a re-run of spring 2018's 0305/012 experiment but using the current version of genTrans.pl which had made slight changes from a year ago.

Note that the filler words [NOISE], [LAUGHTER], [VOCALIZED-NOISE] are dealt with differently in the current genTrans.pl as it leaves all of them in the training transcripts whereas in the spring '18 version of it, some of the filler words (but not all) were erroneously removed thus causing unreliable results in spring 2018's 0305/012's baseline and why we re-ran it here).

Results:

Note the word error rate improved from 34.4% (0305/012) to 33.6%. It is still worse than 0305/011 (WER of 32.8%) and 0305/013 (WER of 32.7%) suggesting that removing either [] or both [] and - may give an improvement. To address this we will re-run all three experiments (i.e. 0305/011, 0305/013 and 0314/004) with 30 hours of data to see if the pattern continues (note that 0314/009 will be the 30 hour repeat of this experiment).


                 SYSTEM SUMMARY PERCENTAGES by SPEAKER

  ,-----------------------------------------------------------------.
  |                            hyp.trans                            |
  |-----------------------------------------------------------------|
  | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err |
  |=================================================================|
  | Sum/Avg | 4172  60569 | 73.1   18.9    8.0    6.7   33.6   88.4 |
  |=================================================================|
  |  Mean   |  1.3   19.2 | 76.0   18.0    6.0   15.1   39.1   88.8 |
  |  S.D.   |  0.5   16.5 | 18.0   15.1    7.8   28.9   31.9   29.3 |
  | Median  |  1.0   15.0 | 76.2   16.7    3.1    3.4   33.3  100.0 |
  `-----------------------------------------------------------------'