Named Entity Recognition on MSRA standard (test)
95F1 ScoreERNIE 2.0
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ERNIE 2.0Model Version=ERNIE 2.0, Model Scale=LARGE2019.07 | 95 | — | — | |
| ERNIE2019.04 | 93.8 | — | — | |
| ERNIE 1.0Model Version=ERNIE 1.0, Model Scale=BASE2019.07 | 93.8 | — | — | |
| ERNIE 2.0Model Version=ERNIE 2.0, Model Scale=BASE2019.07 | 93.8 | — | — | |
| Lattice LSTMmode=lattice-based2018.05 | 93.18 | 93.57 | 92.79 | |
| BERT2019.04 | 92.6 | — | — | |
| BERTModel Version=BERT, Model Scale=BASE2019.07 | 92.6 | — | — | |
| Char baseline + bichar + softwordmode=character-based, extra_features=bichar + softword2018.05 | 91.87 | 92.97 | 90.8 | |
| Zhang et al.rich hand-crafted features=true2018.05 | 91.18 | 92.2 | 90.18 | |
| Dong et al.architecture=neural LSTM-CRF, features=radical features2018.05 | 90.95 | 91.28 | 90.62 | |
| Zhou et al.2018.05 | 90.28 | 91.86 | 88.75 | |
| Word baseline + char + bichar LSTMmode=word-based, extra_features=char + bichar LSTM2018.05 | 90.28 | 91.05 | 89.53 | |
| Char baselinemode=character-based2018.05 | 88.81 | 90.74 | 86.96 | |
| Lu et al.character embedding features=true2018.05 | 87.94 | — | — | |
| Word baselinemode=word-based2018.05 | 86.65 | 90.57 | 83.06 | |
| Chen et al.2018.05 | 86.2 | 91.22 | 81.71 |