Speech:Exps 0259 002

Description
Author: Justin Thibeault Purpose: Reduce dependence on experiment 0257 by recreating Language Model and checking decode with new Language Model. Details: First I'm going to reduce dependence on experiment 0257 by reproducing 0257's language model. sed -e :a -e 's/<[^>]*>//g;/ noaa_full_parsed cp -i /mnt/main/scripts/user/lm_create.pl. ./lm_create noaa_full_parsed mv tmp.lm.DMP NOAA_full.lm.DMP

Then I start the decode by running the script in the 002 subfolder.

Results
SYSTEM SUMMARY PERCENTAGES by SPEAKER

,---.  |                         noaaFull002NewLM.txt                          | |---|  | SPKR         | # Snt  # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |--+--+-|  | noaa_162.450 |  185    6640 | 55.4    9.1   35.5    0.3   45.0  100.0 | |=======================================================================|  | Sum/Avg      |  185    6640 | 55.4    9.1   35.5    0.3   45.0  100.0 | |=======================================================================|  |     Mean     |185.0  6640.0 | 55.4    9.1   35.5    0.3   45.0  100.0 | |    S.D.     |  0.0    0.0  |  0.0    0.0    0.0    0.0    0.0    0.0 | |   Median    |185.0  6640.0 | 55.4    9.1   35.5    0.3   45.0  100.0 | `---'

Analysis
Slight improvement of results over 001 due to Marcel's corrections in the transcript.