Nested Named Entity Recognition on GENIA
83.8F1 ScoreBERT-MRC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BERT-MRCAdditional Supervision=true2020.12 | 83.8 | 85.2 | 81.1 | |
| BERT-MRC2019.10 | 83.75 | 85.18 | 81.12 | |
| Triaffine2026.01 | 81 | 80 | 82 | |
| Yu, Bohnet, and PoesioAdditional Supervision=true2020.12 | 80.5 | 81.8 | 79.3 | |
| BARTNER (Span)Backbone=BART-Large, Entity Representation=Span2021.06 | 79.23 | 78.87 | 79.6 | |
| PyramidEncoder=BERT2020.12 | 79.2 | 79.5 | 78.9 | |
| Wang et al. (2020a)Backbone=BERT-Large, Configuration=Fair comparison2021.06 | 79.19 | 79.45 | 78.94 | |
| Span-level Graphs2026.01 | 79 | 77 | 80 | |
| BARTNER (Word)Backbone=BART-Large, Entity Representation=Word2021.06 | 78.93 | 78.57 | 79.3 | |
| Yu et al. (2020)Backbone=BERT-Large, Reproduction=Sentence-level context only2021.06 | 78.87 | 79.43 | 78.32 | |
| Li et al. (2020b)Backbone=BERT-Large, Reproduction=Rerun code2021.06 | 78.72 | 81.25 | 76.36 | |
| BARTNER (BPE)Backbone=BART-Large, Entity Representation=BPE2021.06 | 78.6 | 78.15 | 79.06 | |
| Straková et al. (2019)Word Embeddings=BERT+FLAIR2019.09 | 78.31 | 80.11 | 76.6 | |
| Seq2seq-BERT2019.10 | 78.31 | — | — | |
| BENSCEncoder=BERT2020.12 | 78.3 | 79.2 | 77.4 | |
| Straková et al. (2019)Word Embeddings=BERT, Use POS tags=true2019.09 | 78.2 | 79.92 | 76.55 | |
| Seq2seqEncoder=Pretrained LM2020.12 | 78.2 | — | — | |
| PO-TreeCRFsEncoder=BERT2020.12 | 78.2 | 78.2 | 78.2 | |
| PO-TreeCRFs2026.01 | 78 | 78 | 78 | |
| PyramidEncoder=None2020.12 | 77.8 | 78.6 | 77 | |
| PO-TreeCRFsEncoder=BERT, Ablation=Change Biaffine to Bilinear2020.12 | 77.6 | 79.9 | 75.5 | |
| PO-TreeCRFsEncoder=BERT, Ablation=W/o. Structure Smoothing2020.12 | 77.6 | 78.7 | 76.5 | |
| PO-TreeCRFsEncoder=BERT, Ablation=W/o. TreeCRFs2020.12 | 77.4 | 80.5 | 74.5 | |
| Second-best Sequence Learning and DecodingWord Embeddings=BERT+FLAIR2019.09 | 77.36 | 77.81 | 76.94 | |
| PO-TreeCRFsEncoder=BERT, Ablation=W/o. Potential Normalization and Structure Smoothing2020.12 | 77.3 | 76.5 | 78.1 | |
| Path-BERT2019.10 | 77.25 | 78.07 | 76.45 | |
| Second-best Sequence Learning and DecodingWord Embeddings=None2019.09 | 77.19 | 78.7 | 75.74 | |
| Sohrab and Miwa (2018)Word Embeddings=None2019.09 | 77.1 | 93.2 | 64 | |
| Second-best Sequence Learning and DecodingWord Embeddings=BERT2019.09 | 77.05 | 77.46 | 76.65 | |
| Shibuya and Hovy (2020)Backbone=BERT-Large, Reproduction=Rerun code2021.06 | 77.05 | 77.46 | 76.65 | |
| BERT-CRFoptimization=Bayesian Optimization2026.01 | 77 | 77 | 77 | |
| Straková et al. (2019)Word Embeddings=None, Use POS tags=true2019.09 | 76.44 | 79.6 | 73.53 | |
| Straková et al. (2019)Backbone=BERT-Large2021.06 | 76.44 | — | — | |
| DYGIEAdditional Supervision=true2020.12 | 76.2 | — | — | |
| Luan et al. (2019)Backbone=ELMo2021.06 | 76.2 | — | — | |
| DYGIE2019.10 | 76.2 | — | — | |
| Biaffine2026.01 | 76 | 79 | 73 | |
| BENSCEncoder=None2020.12 | 75.7 | 78.9 | 72.7 | |
| Wang and Lu (2018)Word Embeddings=None, Use POS tags=true2019.09 | 75.1 | 77 | 73.3 | |
| SHEncoder=None, c=n2020.12 | 75.1 | 77 | 73.3 | |
| Pyramid2026.01 | 75 | 78 | 72 | |
| Recursive-CRF2026.01 | 75 | 75 | 75 | |
| MLC2026.01 | 75 | 77 | 74 | |
| Lin et al. (2019)Word Embeddings=None, Use POS tags=true2019.09 | 74.8 | 75.8 | 73.9 | |
| ARN2019.10 | 74.8 | 75.8 | 73.9 | |
| Ju et al. (2018)Word Embeddings=None2019.09 | 74.7 | 78.5 | 71.3 | |
| Zheng et al. (2019)Word Embeddings=None2019.09 | 74.7 | 75.9 | 73.6 | |
| Cascaded-CRFEncoder=None2020.12 | 74.7 | 78.5 | 71.3 | |
| Hyper-Graph2019.10 | 74.6 | 77.7 | 71.8 | |
| Wang et al. (2018)Word Embeddings=None, Use POS tags=true2019.09 | 73.9 | 78 | 70.2 | |
| TransitionEncoder=None2020.12 | 73.9 | 78 | 70.2 | |
| Katiyar and Cardie (2018)Word Embeddings=None2019.09 | 73.6 | 79.8 | 68.2 | |
| FOFEEncoder=None, c=62020.12 | 71.4 | 75.4 | 67.8 | |
| Layered2026.01 | 71 | 73 | 68 | |
| Exhaustive2026.01 | 71 | 74 | 69 | |
| LSTM-CRFEncoder=None2020.12 | 69.5 | 75.2 | 64.6 |