Timeshrinking
From SpeechWiki
(Difference between revisions)
| Line 2: | Line 2: | ||
{| class="wikitable" | {| class="wikitable" | ||
| - | ! <math>\tau</math> !! dev 2k utt WER !! frames dropped | + | ! train <math>\tau</math> !! test <math>\tau</math> !! dev 2k utt WER !! frames dropped |
|- | |- | ||
| - | | .6 || 69.3% || ~35% | + | | .6 || .6 || 69.3% || ~35% |
|- | |- | ||
| - | | .9 || xx || ~5% | + | | .9 || .9 || [{{SERVER}}/speech/akantor/fisher/exp/timeshrink/test/unit.tri.timeshrink.point9/LATEST.log 56.3%] || ~5% |
| + | |- | ||
| + | | 1 (old baseline)|| .9 || [{{SERVER}}/speech/akantor/fisher/exp/timeshrink/testOnBaseline/unit.tri.timeshrink.point9/LATEST.log 53.9%]|| ~5% | ||
| + | |- | ||
| + | | .9 || 1 || xx || ~5% | ||
|} | |} | ||
| - | I have to check for bugs. It could be that the threshold is too low or it could be something else too. | + | I have to check for bugs. It could be that the threshold is too low or it could be something else too. We should probably rerun baseline too, just to make sure I didn't optimize it unfairly |
Revision as of 17:27, 15 September 2009
Fisher experiments
| train <math>\tau</math> | test <math>\tau</math> | dev 2k utt WER | frames dropped |
|---|---|---|---|
| .6 | .6 | 69.3% | ~35% |
| .9 | .9 | 56.3% | ~5% |
| 1 (old baseline) | .9 | 53.9% | ~5% |
| .9 | 1 | xx | ~5% |
I have to check for bugs. It could be that the threshold is too low or it could be something else too. We should probably rerun baseline too, just to make sure I didn't optimize it unfairly
Future Directions
- Can be viewed as a two-mode special case of best-first viterbi search. So make a real best-first lattice search. Mark mentioned some attempts in the 80'ies to do this.
