Part-of-Speech Tagging on UD Indo-European 1.2 (test)
96.63Accuracybi-LSTM (w+c+Polyglot)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| bi-LSTM (w+c+Polyglot)Input Representation=Words + Characters, Pre-trained Embeddings=Polyglot2016.04 | 96.63 | 82.77 | |
| FREQBINInput Representation=Words + Characters, Pre-trained Embeddings=Polyglot, Training Strategy=Multi-task2016.04 | 96.63 | 87.63 | |
| bi-LSTM (w)Input Representation=Words2016.04 | 96.15 | — | |
| TNTInput Representation=Words2016.04 | 94.7 | — | |
| CRFInput Representation=Words2016.04 | 94.58 | — | |
| bi-LSTM (c)Input Representation=Characters2016.04 | 94.58 | — | |
| bi-LSTM (b)Input Representation=Bytes2016.04 | 94.28 | — | |
| bi-LSTM (d)Input Representation=d2016.04 | 92.72 | — |