Inference speed (sentences/s) on Weibo (Named Entity Recognition)
6.1Throughput (sentences/s)SoftLexicon (LSTM)
Evaluation Results
| Method | Links | |
|---|---|---|
| SoftLexicon (LSTM)Batch size=1, GPU=NVIDIA TITAN X2019.08 | 6.1 | |
| SoftLexicon (LSTM) + bicharBatch size=1, GPU=NVIDIA TITAN X, Features=bichar2019.08 | 5.91 | |
| BERT-taggerBatch size=1, GPU=NVIDIA TITAN X2019.08 | 4.45 | |
| SoftLexicon (LSTM) + BERTBatch size=1, GPU=NVIDIA TITAN X, Backbone=BERT2019.08 | 2.85 | |
| BERT + LSTM + CRFBatch size=1, GPU=NVIDIA TITAN X2019.08 | 2.84 | |
| LR-CNNBatch size=1, GPU=NVIDIA TITAN X2019.08 | 2.41 | |
| Lattice-LSTMBatch size=1, GPU=NVIDIA TITAN X2019.08 | 1 |