Speech:Exps 0281 005

Description
Author: Ben Leith Date: 2/3/16

Purpose: Run any successful train using the current Spring '16 instructions and manual experiment creation method (non-scripted)

Details: First successful train and decode, scoring fails on execution, possibly will require the experiment to be replicated. Without decent scoring, there's no way to know how well the experiment actually went, or if it indeed went at all.

Results: Train execution successful. Language Model Creation Successful. Decode successful. Scoring fails with "Segmentation Fault". I'll need to look into why this is. Perhaps parseDecode isn't parsing correctly and sclite is confused by its input.

Output (to console instead of file for sharing here): [bgl1@caesar etc]$ sclite -r 005_train.trans -h hyp.trans -i swb sclite: 2.3 TK Version 1.3 Begin alignment of Ref File: '005_train.trans' and Hyp File: 'hyp.trans' SYSTEM SUMMARY PERCENTAGES by SPEAKER Segmentation fault (core dumped)