Nested Named Entity Recognition on ACE 2005
86.9F1 ScoreBERT-MRC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BERT-MRCAdditional Supervision=true2020.12 | 86.9 | 87.2 | 86.6 | |
| BERT-MRC2019.10 | 86.88 | 87.16 | 86.59 | |
| Yu, Bohnet, and PoesioAdditional Supervision=true2020.12 | 85.4 | 85.2 | 85.6 | |
| PO-TreeCRFsEncoder=BERT2020.12 | 85.4 | 84.5 | 86.4 | |
| BARTNER (Word)Backbone=BART-Large, Entity Representation=Word2021.06 | 84.74 | 83.16 | 86.38 | |
| PyramidEncoder=BERT2020.12 | 84.7 | 84 | 85.4 | |
| PO-TreeCRFsEncoder=BERT, Ablation=Change Biaffine to Bilinear2020.12 | 84.7 | 83 | 86.5 | |
| Wang et al. (2020a)Backbone=BERT-Large, Configuration=Fair comparison2021.06 | 84.66 | 83.95 | 85.39 | |
| Yu et al. (2020)Backbone=BERT-Large, Reproduction=Sentence-level context only2021.06 | 84.61 | 84.5 | 84.72 | |
| PO-TreeCRFsEncoder=BERT, Ablation=W/o. Structure Smoothing2020.12 | 84.6 | 83.5 | 85.8 | |
| Li et al. (2020b)Backbone=BERT-Large, Reproduction=Rerun code2021.06 | 84.57 | 85.01 | 84.13 | |
| PO-TreeCRFsEncoder=BERT, Ablation=W/o. Potential Normalization and Structure Smoothing2020.12 | 84.4 | 82.7 | 86.2 | |
| Second-best Sequence Learning and DecodingWord Embeddings=BERT+FLAIR2019.09 | 84.34 | 83.83 | 84.87 | |
| Straková et al. (2019)Word Embeddings=BERT+FLAIR2019.09 | 84.33 | 83.48 | 85.21 | |
| Seq2seq-BERT2019.10 | 84.33 | — | — | |
| PO-TreeCRFsEncoder=BERT, Ablation=W/o. TreeCRFs2020.12 | 84.1 | 82 | 86.4 | |
| Second-best Sequence Learning and DecodingWord Embeddings=BERT2019.09 | 83.99 | 83.3 | 84.69 | |
| Shibuya and Hovy (2020)Backbone=BERT-Large, Reproduction=Rerun code2021.06 | 83.99 | 83.3 | 84.69 | |
| BENSCEncoder=BERT2020.12 | 83.9 | 83.8 | 83.9 | |
| Straková et al. (2019)Word Embeddings=BERT, Use POS tags=true2019.09 | 83.42 | 82.58 | 84.29 | |
| Straková et al. (2019)Backbone=BERT-Large2021.06 | 83.42 | — | — | |
| Seq2seqEncoder=Pretrained LM2020.12 | 83.4 | — | — | |
| DYGIEAdditional Supervision=true2020.12 | 82.9 | — | — | |
| Luan et al. (2019)Backbone=ELMo2021.06 | 82.9 | — | — | |
| DYGIE2019.10 | 82.9 | — | — | |
| BARTNER (BPE)Backbone=BART-Large, Entity Representation=BPE2021.06 | 82.75 | 82.08 | 83.44 | |
| Path-BERT2019.10 | 82.7 | 82.98 | 82.42 | |
| Fisher and Vlachos (2019)Word Embeddings=BERT2019.09 | 82.4 | 82.7 | 82.1 | |
| MLEncoder=BERT2020.12 | 82.4 | 82.7 | 82.1 | |
| Merge-BERT2019.10 | 82.4 | 82.7 | 82.1 | |
| BARTNER (Span)Backbone=BART-Large, Entity Representation=Span2021.06 | 82.31 | 81.41 | 83.24 | |
| PyramidEncoder=None2020.12 | 79.4 | 80 | 78.9 | |
| MLEncoder=ELMo2020.12 | 78.9 | 79.7 | 78 | |
| MGNEREncoder=ELMo2020.12 | 78.2 | 79 | 77.3 | |
| Second-best Sequence Learning and DecodingWord Embeddings=None2019.09 | 76.83 | 78.27 | 75.44 | |
| BENSCEncoder=None2020.12 | 75.6 | 77.1 | 74.2 | |
| Straková et al. (2019)Word Embeddings=None, Use POS tags=true2019.09 | 75.36 | 76.35 | 74.39 | |
| Lin et al. (2019)Word Embeddings=None, Use POS tags=true2019.09 | 74.9 | 76.2 | 73.6 | |
| ARN2019.10 | 74.9 | 76.2 | 73.6 | |
| Fisher and Vlachos (2019)Word Embeddings=None2019.09 | 74.6 | 75.1 | 74.1 | |
| MLEncoder=None2020.12 | 74.6 | 75.1 | 74.1 | |
| Wang and Lu (2018)Word Embeddings=None, Use POS tags=true2019.09 | 74.5 | 76.8 | 72.3 | |
| SHEncoder=None, c=n2020.12 | 74.5 | 76.8 | 72.3 | |
| Seg-Graph2019.10 | 74.5 | 76.8 | 72.3 | |
| Wang et al. (2018)Word Embeddings=None, Use POS tags=true2019.09 | 73 | 74.5 | 71.5 | |
| TransitionEncoder=None2020.12 | 73 | 74.5 | 71.5 | |
| Ju et al. (2018)Word Embeddings=None2019.09 | 72.2 | 74.2 | 70.3 | |
| Cascaded-CRFEncoder=None2020.12 | 72.2 | 74.2 | 70.3 | |
| FOFEEncoder=None, c=62020.12 | 71 | 76.5 | 66.3 | |
| Katiyar and Cardie (2018)Word Embeddings=None2019.09 | 70.5 | 70.6 | 70.4 | |
| Hyper-Graph2019.10 | 70.5 | 70.6 | 70.4 | |
| LSTM-CRFEncoder=None2020.12 | 62.2 | 70.3 | 55.7 |