POS tagging on UD Treebank English 2.15
91.9AccuracyConditional Random Fields (CRF)
Evaluation results
| Method | Links | |
|---|---|---|
| Conditional Random Fields (CRF)Encoder layers=2, Training approach=CRF training, Decoding strategy=Viterbi2025.05 | 91.9 | |
| Bregman CRF (BCRF)Encoder layers=2, Training approach=CRF training, Decoding strategy=BCRF, Iterations=52025.05 | 91.9 | |
| Bregman CRF (BCRF)Encoder layers=2, Training approach=CRF training, Decoding strategy=BCRF, Iterations=102025.05 | 91.9 | |
| Bregman CRF (BCRF)Encoder layers=2, Training approach=BCRF training, Decoding strategy=Viterbi2025.05 | 91.8 | |
| Bregman CRF (BCRF)Encoder layers=2, Training approach=BCRF training, Decoding strategy=BCRF, Iterations=52025.05 | 91.8 | |
| Bregman CRF (BCRF)Encoder layers=2, Training approach=BCRF training, Decoding strategy=BCRF, Iterations=102025.05 | 91.8 | |
| Mean Field (MF)Encoder layers=2, Training approach=MF training, Decoding strategy=MF, Iterations=52025.05 | 91.7 | |
| Mean Field (MF)Encoder layers=2, Training approach=MF training, Decoding strategy=MF, Iterations=102025.05 | 91.7 | |
| Mean Field (MF)Encoder layers=2, Training approach=MF training, Decoding strategy=Viterbi2025.05 | 91.6 | |
| Mean Field (MF)Encoder layers=0, Training approach=MF training, Decoding strategy=MF, Iterations=52025.05 | 91.4 | |
| Mean Field (MF)Encoder layers=0, Training approach=MF training, Decoding strategy=MF, Iterations=102025.05 | 91.4 | |
| Conditional Random Fields (CRF)Encoder layers=0, Training approach=CRF training, Decoding strategy=Viterbi2025.05 | 91.3 | |
| Bregman CRF (BCRF)Encoder layers=0, Training approach=CRF training, Decoding strategy=BCRF, Iterations=52025.05 | 91.3 | |
| Bregman CRF (BCRF)Encoder layers=0, Training approach=CRF training, Decoding strategy=BCRF, Iterations=102025.05 | 91.3 | |
| Bregman CRF (BCRF)Encoder layers=0, Training approach=BCRF training, Decoding strategy=Viterbi2025.05 | 91.1 | |
| Bregman CRF (BCRF)Encoder layers=0, Training approach=BCRF training, Decoding strategy=BCRF, Iterations=52025.05 | 91.1 | |
| Bregman CRF (BCRF)Encoder layers=0, Training approach=BCRF training, Decoding strategy=BCRF, Iterations=102025.05 | 91.1 | |
| Mean Field (MF)Encoder layers=2, Training approach=CRF training, Decoding strategy=MF, Iterations=52025.05 | 91 | |
| Mean Field (MF)Encoder layers=2, Training approach=CRF training, Decoding strategy=MF, Iterations=102025.05 | 91 | |
| Unstructured BaselineEncoder layers=2, Training approach=Baseline2025.05 | 90.8 | |
| Mean Field (MF)Encoder layers=0, Training approach=CRF training, Decoding strategy=MF, Iterations=102025.05 | 90.7 | |
| Mean Field (MF)Encoder layers=0, Training approach=CRF training, Decoding strategy=MF, Iterations=52025.05 | 90.6 | |
| Mean Field (MF)Encoder layers=0, Training approach=MF training, Decoding strategy=Viterbi2025.05 | 89.1 | |
| Unstructured BaselineEncoder layers=0, Training approach=Baseline2025.05 | 85.8 |