Tagger models
Result legend
Evaluation results for the evaluation models are reported in the following format:
Experiment statistics: GENRE:MODEL:OBSOLETE_INFO *Overall: mean=95.91 std=0.15 min=95.67 max=96.10 *Known: mean=96.33 std=0.15 min=96.13 max=96.55 *Unknown: mean=83.80 std=1.01 min=81.62 max=85.03 *Proportion unknown: mean=3.35 std=0.24 min=2.95 max=3.66 /.../ Best combination: GENRE:MODEL:OBSOLETE_INFO
where
- GENRE is either all or one of the SUC genres (a-k)
- MODEL is the current model evaluated
- OBSOLETE_INFO is residual info from older runs (ignore)
- Overall is overall accuracy (mean, standard deviation, min, and max)
- Known is accuracy for known words (mean, standard deviation, min, and max)
- Unknown is accuracy for unknown words (mean, standard deviation, min, and max)
- Proportion unknown is the proportion of unknown words (mean, standard deviation, min, and max)
- Best combination is residual info from older runs (ignore)
The accuracy and proportion results are computed with tnt-diff on 10 folds of SUC, using a TnT lexical model trained on the same material as the evaluated tagger to get information on which words are known or unknown, while the standard deviation, min and max results are computed with the Perl module Statistics::Descriptive.