Entity Linking on AQUAINT (test)
93.7Micro F1 ScoreOracle type (Ultra-fine)
Evaluation results
| Method | Links | |
|---|---|---|
| Oracle type (Ultra-fine)Type source=Ultra-fine, Mode=Oracle2020.01 | 93.7 | |
| Oracle type (FIGER)Type source=FIGER, Mode=Oracle2020.01 | 92.3 | |
| RI2019.09 | 90 | |
| Cheng and Roth (2013)2020.01 | 90 | |
| BERT-Entity-Sim (local & global)2020.01 | 89.8 | |
| Predict type (Ultra-fine)Type source=Ultra-fine, Mode=Predicted2020.01 | 89.7 | |
| Predict type (ZOE)Type source=ZOE, Mode=Predicted2020.01 | 89.5 | |
| BERT+G&H's embeddings2020.01 | 89.1 | |
| Berkeley-CNN + DCA-RLBackbone=Berkeley-CNN, Learning protocol=Reinforcement Learning2019.09 | 88.53 | |
| Deep-ED2019.09 | 88.5 | |
| Ganea and Hofmann (2017)2020.01 | 88.5 | |
| Ment-Norm2019.09 | 88.3 | |
| Le and Titov (2018)2020.01 | 88.3 | |
| ETHZ-Attn + DCA-RLBackbone=ETHZ-Attn, Learning protocol=Reinforcement Learning2019.09 | 88.25 | |
| ETHZ-Attn + DCA-SLBackbone=ETHZ-Attn, Learning protocol=Supervised Learning2019.09 | 87.38 | |
| WNED2019.09 | 87 | |
| Guo and Barbosa (2016)2020.01 | 87 | |
| Berkeley-CNN + DCA-SLBackbone=Berkeley-CNN, Learning protocol=Supervised Learning2019.09 | 85.63 | |
| Milne and Witten (2008)2020.01 | 85 | |
| ETHZ-AttnBackbone=ETHZ-Attn2019.09 | 84.06 | |
| Prior (p(e|m))2019.09 | 83.2 | |
| prior p(e|m)2020.01 | 83.2 | |
| GLOW2019.09 | 83 | |
| Ratinov et al. (2011)2020.01 | 83 | |
| Berkeley-CNNBackbone=Berkeley-CNN2019.09 | 80.55 | |
| AIDA2019.09 | 56 | |
| Hoffart et al. (2011)2020.01 | 56 |