Speech:Exps 0304 017

Description
Author: Stephen Thibault (UserID: sdt1001)

Date: 3-6-2018

Purpose: Data from 0301/006

Details: Copied to this sub-experiment the data from 0301/006 which is the data train used in the Unseen Decoding and Scoring in 0301/011 which achieved a 28.4% WER.

Results: Copying completed. Renamed all files with "006" with "017" so scripts will not error out in the Decode process.

NOTE: 0301/006 did NOT have a Language Model present when the sub-experiment contents were copied to this sub-experiment directory. This will allow us to use a different LM (with input provided from the 2018 Data Group) for Decoding. The 0301/011 has the LM present so they could have created it there and then Decoded and Scored or maybe it was created in /006 and then deleted, but we don't know.

ABANDONED. Determined on 17 April that just activating LDA in the sphinx_train.cfg file is NOT the only necessity to train with LDA correctly. So the results of 28.4% WER obtained by 2017 Spring Rebels Team in 0301/011 are irrelevant.