Speech:Exps 0284 001

Description
Author: Brian Anker Date: 3/14/2016

Purpose: Learn the training process

Details: .

Results: 3/14: Was able to run a 5 hr train from Asterix. The script paths from /mnt/main/script/user/ are not mapped to asterix at the moment. 3/15: Successfully ran a train on first_5hr corpus, built a language model from the first 3000 files(3506 total files in 001_Train.fileids), and decoded on the trained data. Below are my results from the train. SYSTEM SUMMARY PERCENTAGES by SPEAKER ,-.    |                            hyp.trans                            | |-|    | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |-+-+-|    | sw2001b |   18    163 | 79.8   16.0    4.3   39.9   60.1  100.0 | |-+-+-|    | sw2001a |   14    101 | 74.3   23.8    2.0   52.5   78.2  100.0 | |-+-+-|    | sw2005a |   13    276 | 80.4   14.9    4.7   13.4   33.0  100.0 | |-+-+-|    | sw2005b |   27    307 | 61.6   22.5   16.0   16.0   54.4  100.0 | |=================================================================|    | Sum/Avg |   72    847 | 72.7   18.9    8.4   24.1   51.4  100.0 | |=================================================================|    |  Mean   | 18.0  211.8 | 74.0   19.3    6.7   30.4   56.4  100.0 | | S.D.   |  6.4   96.3 |  8.7    4.5    6.3   18.9   18.6    0.0 | | Median | 16.0  219.5 | 77.0   19.2    4.5   27.9   57.3  100.0 | `-'