Named Entity Recognition on ACE 2005 (test)
89.9F1 ScoreHySPA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HySPASpace Complexity=O(n), Time Complexity=O(n), Backbone=ALBERT2021.06 | 89.9 | — | — | |
| Table-Sequence EncodersAveraging=micro-averaged2020.10 | 89.5 | — | — | |
| TabSeqSpace Complexity=O(n²), Time Complexity=O(n), Backbone=ALBERT-xxlarge2021.06 | 89.5 | — | — | |
| HySPASpace Complexity=O(n), Time Complexity=O(n), Backbone=RoBERTa2021.06 | 88.9 | — | — | |
| OneIESpace Complexity=O(n), Time Complexity=O(n²)2021.06 | 88.8 | — | — | |
| TablERTSentence=Multi2020.10 | 88.7 | 88.8 | 88.6 | |
| Wadden et al.Averaging=micro-averaged2020.10 | 88.6 | — | — | |
| Dygie++Space Complexity=O(n), Time Complexity=O(n²)2021.06 | 88.6 | — | — | |
| DyGIE++Sentence=Multi2020.10 | 88.6 | — | — | |
| Luan et al.Averaging=micro-averaged2020.10 | 88.4 | — | — | |
| TablERTSentence=Single2020.10 | 87.6 | 87.2 | 88.1 | |
| Baseline + BSboundary smoothing=true2022.04 | 87.15 | 86.25 | 88.07 | |
| Li et al.source=2020b2022.04 | 86.88 | 87.16 | 86.59 | |
| W2NERMethod Category=Proposed Method2021.12 | 86.79 | 85.03 | 88.62 | |
| Shen et al.Method Category=Span-based2021.12 | 86.67 | 86.09 | 87.27 | |
| Shen et al.2022.04 | 86.67 | 86.09 | 87.27 | |
| Baseline2022.04 | 86.56 | 84.29 | 88.97 | |
| Dixit and AlAveraging=micro-averaged2020.10 | 86 | — | — | |
| SpanRESpace Complexity=O(n), Time Complexity=O(n²)2021.06 | 86 | — | — | |
| Yu et al.Method Category=Span-based2021.12 | 85.4 | 85.2 | 85.6 | |
| Yu et al.2022.04 | 85.4 | 85.2 | 85.6 | |
| Li et al.Averaging=micro-averaged2020.10 | 84.8 | — | — | |
| Yan et al.Method Category=Seq2Seq2021.12 | 84.74 | 83.16 | 86.38 | |
| Wang et al.Method Category=Span-based2021.12 | 84.66 | 83.95 | 85.39 | |
| Straková et al.trained with dev split=true2022.04 | 84.33 | — | — | |
| Seq2Seq, ELMo, BERT, FlairReference=Straková et al. (2019)2021.06 | 84.3 | — | — | |
| Zhang et al.Averaging=micro-averaged2020.10 | 83.6 | — | — | |
| Sun et al.Averaging=micro-averaged2020.10 | 83.6 | — | — | |
| Straková et al.Method Category=Seq2Seq2021.12 | 83.42 | — | — | |
| Miwa and BansalAveraging=micro-averaged2020.10 | 83.4 | — | — | |
| Span-based, BERT (Our Model)Embeddings=BERT, Architecture=Span-based2021.06 | 83 | — | — | |
| Span-based, ELMo, CorefReference=Luan et al. (2019)2021.06 | 82.9 | — | — | |
| Span-based model (- Overlap Relation)Ablation=w/o Overlap Relation2021.06 | 82.7 | — | — | |
| Katiyar and CardieAveraging=micro-averaged2020.10 | 82.6 | — | — | |
| PointerNetSpace Complexity=O(n), Time Complexity=O(n²)2021.06 | 82.6 | — | — | |
| Span-based model (- Dep guided GCN)Ablation=w/o Dependency-guided GCN2021.06 | 82.6 | — | — | |
| Merge and label, BERTReference=Fisher and Vlachos (2019)2021.06 | 82.4 | — | — | |
| Li and JiAveraging=micro-averaged2020.10 | 80.8 | — | — | |
| Multi-grained detection, RNN, ELMoReference=Xia et al. (2019)2021.06 | 78.2 | — | — | |
| GEANNGazetteer=true2020.05 | 75.2 | 77.1 | 73.3 | |
| BiFlaG2020.05 | 75.1 | 75 | 75.2 | |
| Anchor-RegionPOS tags=true2020.05 | 74.9 | 76.2 | 73.6 | |
| Anchor-region detection, RNN, CNNReference=Lin et al. (2019a)2021.06 | 74.9 | — | — | |
| Merge & Label2020.05 | 74.6 | 75.1 | 74.1 | |
| Segm. hypPOS tags=true2020.05 | 74.5 | 76.8 | 72.3 | |
| Hypergraph, RNNReference=Wang and Lu (2018)2021.06 | 74.5 | — | — | |
| Wang and LuMethod Category=Hypergraph-based2021.12 | 74.5 | 76.8 | 72.3 | |
| Transition-based parsing, RNNReference=Wang et al. (2018)2021.06 | 73 | — | — | |
| layered-CRF2020.05 | 72.2 | 74.2 | 70.3 | |
| Dynamically stacking, RNNReference=Ju et al. (2018)2021.06 | 72.2 | — | — | |
| Ju et al.Method Category=Sequence Labeling2021.12 | 72.2 | 74.2 | 70.3 | |
| LSTM. hyp2020.05 | 70.5 | 70.6 | 70.4 | |
| Hypergraph, RNNReference=Katiyar and Cardie (2018)2021.06 | 70.5 | — | — | |
| Katiyar and Cardie2022.04 | 70.5 | 70.6 | 70.4 | |
| Multi-CRF2020.05 | 65.2 | 69.7 | 61.3 | |
| LSTM-CRF2020.05 | 62.2 | 70.3 | 55.7 | |
| HypergraphReference=Muis and Lu (2017)2021.06 | 61.3 | — | — | |
| HypergraphReference=Lu and Roth (2015)2021.06 | 58.7 | — | — |