Speech:Exps 0302 004

Description
Author: Zachary Dudek

Date: 12-14-2017

Purpose: A test on Obelix

Details:

Results: Obelix was a machine that I had consistent results on compared to Caesar and Asterix up and to this point. Obelix has the same WER as the other two machines a 43.7%. This is using only using the decode on trained data just as the others are using. I do believe that from some other tests that if I had used decode on unseen data I would get more real world results but the results that I derived are just a baseline to determine consistency among the machines. SYSTEM SUMMARY PERCENTAGES by SPEAKER

,-.     |                            hyp.trans                            | |-|     | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |-+-+-|     |=================================================================|      | Sum/Avg | 3992  57805 | 63.9   26.8    9.2    7.6   43.7   90.4 | |=================================================================|     |  Mean   |  1.3   18.7 | 68.2   24.9    6.9   15.7   47.5   90.6 | | S.D.   |  0.5   15.9 | 20.9   17.6    9.1   29.1   32.9   27.3 | | Median |  1.0   15.0 | 66.7   25.0    3.6    4.2   42.9  100.0 | `-'