Part-of-Speech Tagging on UD Average 1.2 (test)
96.65AccuracyAT POS tagger
Evaluation results
| Method | Links | |||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AT POS taggerModel Architecture=BiLSTM-CRF, Training Protocol=Adversarial Training2017.11 | 96.65 | — | — | 98.53 | 96.74 | 94.35 | 95.82 | 96.44 | 94.71 | 97.51 | 95.4 | 96.63 | 97.21 | 94.03 | 98.08 | — | 93.09 | 98.08 | 97.57 | 98.07 | 98.11 | 98.81 | 97.43 | 96.32 | 96.7 | 98.24 | 91.32 | 91.11 | 94.02 | 91.46 | 83.16 | |
| FREQBINInput Representation=Words + Characters, Pre-trained Embeddings=Polyglot, Training Strategy=Multi-task (FreqBin)2016.04 | 96.52 | 87.98 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| bi-LSTM (w+c+Polyglot)Input Representation=Words + Characters, Pre-trained Embeddings=Polyglot2016.04 | 96.5 | 83.48 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BiLSTM-CRF BaselineModel Architecture=BiLSTM-CRF, Training Protocol=Standard2017.11 | 96.45 | — | — | 98.34 | 96.63 | 94.29 | 95.72 | 96.26 | 94.55 | 97.38 | 94.54 | 96.48 | 97.12 | 93.95 | 98.04 | — | 92.64 | 97.88 | 97.34 | 97.94 | 97.81 | 98.7 | 97.34 | 96.12 | 96.39 | 98.18 | 90.79 | 90.66 | 93.39 | 91.24 | 82.91 | |
| Multi-task BiLSTM (Plank et al., 2016)Model Architecture=Multi-task BiLSTM2017.11 | 96.4 | — | — | 97.97 | 96.35 | 93.38 | 95.16 | 95.74 | 95.51 | 97.49 | 95.85 | 96.11 | 97.1 | 93.41 | 97.95 | — | 93.3 | 98.03 | 97.62 | 97.9 | 96.84 | 98.24 | 96.96 | 96.82 | 96.69 | — | — | — | — | — | — | |
| bi-LSTM (w)Input Representation=Words2016.04 | 96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BiLSTM-aux2017.05 | 95.9 | — | 98.9 | 98 | 96.2 | 92.6 | 94.5 | 95.1 | 94.7 | 97.2 | 94.9 | 95.8 | 96.2 | 93.1 | 97.6 | 95.8 | 93.3 | 97.6 | 96.4 | 97.5 | 97.6 | — | — | — | — | — | — | — | — | — | — | |
| BTS2016.04 | 95.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| jPTDP2017.05 | 95.7 | — | 98.8 | 97.4 | 95.8 | 92.7 | 94.7 | 95.9 | 93.7 | 96.8 | 94.6 | 96 | 96.4 | 93.1 | 97.5 | 95.5 | 91.4 | 97.4 | 96.3 | 97.5 | 97.1 | — | — | — | — | — | — | — | — | — | — | |
| Nguyen et al. (2017)2017.11 | 95.55 | — | — | 97.4 | 95.8 | 92.7 | 94.7 | 95.9 | 93.7 | 96.8 | 94.6 | 96 | 96.4 | — | 97.5 | — | 91.4 | 97.4 | 96.3 | 97.5 | 97.1 | 98 | — | — | — | — | — | — | — | — | — | |
| Stack-prop2017.05 | 95.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UDPipe2017.05 | 95.3 | — | 98.7 | 97.8 | 95.8 | 90.7 | 94.5 | 95 | 93.1 | 96.9 | 94.9 | 95.9 | 95.8 | 93.6 | 97.2 | 94.8 | 89.2 | 97.2 | 96 | 97.4 | 95.6 | — | — | — | — | — | — | — | — | — | — | |
| TNTInput Representation=Words2016.04 | 94.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TNT (Plank et al., 2016)Model Architecture=TNT2017.11 | 94.55 | — | — | 96.84 | 94.29 | 92.64 | 94.55 | 94.55 | 93.35 | 95.98 | 93.59 | 94.51 | 94.53 | 93.16 | 96.16 | — | 88.54 | 96.31 | 95.57 | 96.27 | 94.92 | 96.82 | 93.71 | 94.06 | 95.19 | — | — | — | — | — | — | |
| TnT2017.05 | 94.5 | — | 97.8 | 96.8 | 94.3 | 92.6 | 92.7 | 94.6 | 93.4 | 96 | 93.6 | 94.5 | 94.5 | 93.2 | 96.2 | 93.7 | 88.5 | 96.3 | 95.6 | 96.3 | 94.9 | — | — | — | — | — | — | — | — | — | — | |
| bi-LSTM (c)Input Representation=Characters2016.04 | 94.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CRFInput Representation=Words2016.04 | 94.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CRF2017.05 | 94.2 | — | 97.6 | 96.4 | 93.8 | 91.4 | 93.4 | 94.2 | 91.6 | 95.7 | 90.3 | 95.1 | 96 | 93 | 96.4 | 93.6 | 90 | 96.2 | 94 | 96.3 | 94.8 | — | — | — | — | — | — | — | — | — | — | |
| CRF (Plank et al., 2016)Model Architecture=CRF2017.11 | 94.11 | — | — | 96.36 | 93.83 | 91.38 | 93.35 | 94.23 | 91.63 | 95.65 | 90.32 | 95.14 | 96 | 92.96 | 96.43 | — | 90.03 | 96.21 | 93.96 | 96.32 | 94.77 | 96.56 | 93.63 | 93.16 | 94.45 | — | — | — | — | — | — | |
| bi-LSTM (b)Input Representation=Bytes2016.04 | 94.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Berend (2017)2017.11 | 93.59 | — | — | 95.63 | 93.32 | 90.73 | 93.47 | 94.69 | 90.63 | 96.11 | 89.19 | 94.96 | 96.09 | 92.02 | 96.28 | — | 85.1 | 95.67 | 93.95 | 95.5 | 92.7 | 95.83 | 95.28 | 93.53 | 94.62 | 97.12 | 86.3 | 88.82 | 89.47 | 88.99 | 81.8 | |
| bi-LSTM (d)Input Representation=d2016.04 | 92.37 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |