Fine-grained Entity Typing on OpenEntity (test)
80.01PrecisionK-ADAPTER (L)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| K-ADAPTER (L)Knowledge Type=Linguistic (linAdapter)2020.02 | 80.01 | 74 | 76.89 | |
| K-ADAPTER (F)Knowledge Type=Factual (facAdapter)2020.02 | 79.3 | 75.84 | 77.53 | |
| K-ADAPTER (F+L)Knowledge Type=Factual + Linguistic2020.02 | 78.99 | 76.27 | 77.61 | |
| KnowBERT2020.02 | 78.6 | 73.7 | 76.1 | |
| ERNIEKnowledge Injection=Entity-level2020.02 | 78.42 | 72.9 | 75.56 | |
| ROBERTa + multitaskTraining Mode=Multi-task pre-training2020.02 | 77.96 | 76 | 76.97 | |
| ROBERTaBackbone=RoBERTa-base2020.02 | 77.55 | 74.95 | 76.23 | |
| KEPLERKnowledge Injection=Knowledge embedding objective2020.02 | 77.2 | 74.2 | 75.7 | |
| BERT-baseBackbone=BERT-base2020.02 | 76.37 | 70.96 | 73.56 | |
| K-ADAPTER (w/o knowledge)Knowledge Type=None2020.02 | 74.47 | 74.91 | 76.17 | |
| NFGECModel Architecture=Attentive recursive neural networks2020.02 | 68.8 | 53.3 | 60.1 |