Table 2.
184 utterances from hand-coded videos from the Rollins corpus, with parameters fine-tuned on training data. CIs for non-deterministic models calculated by bootstrapping from 1000 resamples of the F-scores.
| Precision | Recall | F1 | |
|---|---|---|---|
| Propose/Verify | 0.03 | 0.50 | 0.06 (95% CI = 0.053–0.057) |
| Fazly et al. 2010 | 0.75 | 0.21 | 0.33 |
| Modified Fazly | 0.25 | 0.43 | 0.32 |
| Pursuit | 0.60 | 0.25 |
0.35 (95% CI = 0.343–0.360) |