End-to-end Relation Extraction on ACE04 (5-fold)
90.3Ent ScorePURE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PUREContext Setting=cross-sentence, Encoder=ALBERT-xxlarge-v12020.10 | 90.3 | 66.1 | 62.2 | |
| PUREContext Setting=cross-sentence, Encoder=BERT-base2020.10 | 89.2 | 63.9 | 60.1 | |
| PUREContext Setting=single-sentence, Encoder=ALBERT-xxlarge-v12020.10 | 88.8 | 64.7 | 60.2 | |
| Wang and LuEncoder=ALBERT-xxlarge-v12020.10 | 88.6 | 63.3 | 59.6 | |
| PUREContext Setting=single-sentence, Encoder=BERT-base2020.10 | 88.1 | 62.8 | 58.3 | |
| Luan et al. (2019)Encoder=LSTM + ELMo, Cross-sentence information usage=true, Additional data usage=coreference2020.10 | 87.4 | 59.7 | — | |
| Li et al.Encoder=BERT-large2020.10 | 83.6 | — | 49.4 | |
| Miwa and BansalEncoder=LSTM2020.10 | 81.8 | — | 48.4 | |
| Li and JiEncoder=LSTM2020.10 | 79.7 | 48.3 | 45.3 | |
| Katiyar and CardieEncoder=LSTM2020.10 | 79.6 | 49.3 | 45.7 |