Speech:Exps 0282 009

Description
Author: Peter Ferro

Date: 3-2-2016

Purpose: This experiment tests the mkDec.pl script file and ensures that it can properly automate the decode process.

Details: This experiment will use the same initial parameters as Speech:Exps 0282 005, only that mkDec.pl will be the one handling the decoding.

Results: mkDec.pl successfully ran the decoder. The results are shown as below. ,-.     |                            hyp.trans                            | |-|     | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |-+-+-|     | sw2001b |   18    163 | 78.5   16.6    4.9   45.4   66.9  100.0 | |-+-+-|     | sw2001a |   14    101 | 81.2   16.8    2.0   57.4   76.2  100.0 | |-+-+-|     | sw2005a |   39    701 | 84.0   10.7    5.3   12.7   28.7   94.9 | |-+-+-|     | sw2005b |   67    613 | 63.0   26.3   10.8   30.0   67.0  100.0 | |-+-+-|     | sw2006b |   29    618 | 72.3   17.8    9.9    7.0   34.6  100.0 | |-+-+-|     | sw2006a |   33    455 | 84.4   12.5    3.1   13.8   29.5  100.0 | |-+-+-|     | sw2007a |   61    614 | 76.7   15.0    8.3   15.8   39.1   88.5 | |-+-+-|     | sw2007b |   60    861 | 81.2   14.8    4.1    8.2   27.1   85.0 | |-+-+-|     | sw2008a |   24    260 | 75.8   20.8    3.5   29.2   53.5  100.0 | |-+-+-|     | sw2008b |   26    257 | 86.8   11.3    1.9   25.7   38.9   92.3 | |-+-+-|     | sw2009b |   23    181 | 76.2   16.0    7.7   29.3   53.0   95.7 | |-+-+-|     | sw2009a |   34    473 | 67.7   27.1    5.3   17.3   49.7  100.0 | |-+-+-|     | sw2010b |   22    404 | 67.3   22.3   10.4   11.6   44.3  100.0 | |-+-+-|     | sw2010a |   27    284 | 77.8   16.9    5.3   20.1   42.3   96.3 | |-+-+-|     | sw2012a |   45    838 | 81.0   13.5    5.5   13.1   32.1   93.3 | |-+-+-|     | sw2012b |   28    464 | 77.4   17.2    5.4   14.9   37.5   96.4 | |-+-+-|     | sw2013a |   35    377 | 67.9   26.5    5.6   21.2   53.3  100.0 | |-+-+-|     | sw2013b |   69    942 | 54.0   30.9   15.1   12.5   58.5  100.0 | |-+-+-|     | sw2014a |    9     79 | 73.4   19.0    7.6   38.0   64.6  100.0 | |-+-+-|     | sw2014b |   13    174 | 80.5   15.5    4.0   17.8   37.4   92.3 | |-+-+-|     | sw2015a |   21    375 | 74.4   13.9   11.7    3.7   29.3   81.0 | |-+-+-|     | sw2015b |   32    542 | 72.3   15.5   12.2    8.9   36.5  100.0 | |-+-+-|     | sw2017b |   38    658 | 81.0   14.1    4.9   14.9   33.9   97.4 | |-+-+-|     | sw2017a |   40    332 | 74.1   22.0    3.9   29.2   55.1   97.5 | |-+-+-|     | sw2018a |   49    505 | 60.6   34.9    4.6   42.2   81.6  100.0 | |-+-+-|     | sw2018b |   42    586 | 71.2   20.8    8.0   20.6   49.5  100.0 | |-+-+-|     | sw2019a |   50    594 | 76.4   17.5    6.1   14.8   38.4   94.0 | |-+-+-|     | sw2019b |   52    452 | 80.5   14.6    4.9   26.3   45.8  100.0 | |=================================================================|     | Sum/Avg | 1000  12903 | 74.0   18.9    7.1   17.8   43.8   96.4 | |=================================================================|     |  Mean   | 35.7  460.8 | 74.9   18.6    6.5   21.5   46.6   96.6 | | S.D.   | 16.3  229.0 |  7.5    5.9    3.3   12.6   14.8    5.0 | | Median | 33.5  459.5 | 76.3   16.9    5.3   17.6   43.3  100.0 | `-'

I sense something unusual, mainly because I ran through the instructions like before, only modifying it to incorporate my mkDec.pl file into it. If switchboard first_5hr/train has been fooled around with, then this would make sense. Otherwise, I am aware of some potential modifications that were taking place via the e-mails I was receiving, so the results may have accidentally been altered between 2/11/16, the date of my first train and 3/5/16, which was when I ran the second train to set up for this script.