Unsupervised Dependency Parsing on WSJ section 23 (all lengths) (test)
65.8Directed Dependency Accuracy (DDA)MaxEnc
Evaluation Results
| Method | Links | |
|---|---|---|
| MaxEncGold POS tags=Yes2018.08 | 65.8 | |
| CS*Gold POS tags=No, Additional punctuation-based constraints=true2018.08 | 64.4 | |
| L-NDMVGold POS tags=Yes, Training scale=Big training data2018.08 | 63.2 | |
| Neural E-DMVGold POS tags=Yes2018.08 | 57.6 | |
| UR-A E-DMVGold POS tags=Yes2018.08 | 57 | |
| CRFAEGold POS tags=Yes2018.08 | 55.7 | |
| Invertible Neural ProjectionsGold POS tags=No, Number of layers=82018.08 | 47.9 | |
| Invertible Neural ProjectionsGold POS tags=No, Number of layers=42018.08 | 46.2 | |
| UR-A E-DMVGold POS tags=No2018.08 | 46.1 | |
| Invertible Neural ProjectionsGold POS tags=No, Number of layers=162018.08 | 43.9 | |
| Gaussian DMVGold POS tags=No2018.08 | 43.1 | |
| Neural E-DMVGold POS tags=No2018.08 | 42.7 | |
| DMVGold POS tags=Yes2018.08 | 39.7 | |
| E-DMVGold POS tags=No2018.08 | 38.2 | |
| DMVGold POS tags=No2018.08 | 35.8 | |
| CRFAEGold POS tags=No2018.08 | 29.5 |