Chunking on CoNLL 2000 (test)
97.3F1 ScoreGCDT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GCDTbackbone=BERT_LARGE2019.06 | 97.3 | — | — | — | |
| Clark et al.2019.06 | 97 | — | — | — | |
| Akbik et al.2019.06 | 96.72 | — | — | — | |
| Peters et al.2017.11 | 96.37 | — | — | — | |
| Peters et al.2019.06 | 96.37 | — | — | — | |
| LD-NetLMs Ind.#=9, Pruning Status=origin2018.04 | 96.15 | — | 51 | — | |
| LM-LSTM-CRFExtra Resource=None, Type=maximum, pre-trained word embedding=true, leverage language models=true2017.09 | 96.13 | — | — | — | |
| LD-NetLMs Ind.#=8, Pruning Status=origin2018.04 | 96.13 | — | 51 | — | |
| Liu et al.2019.06 | 95.96 | — | — | — | |
| Ma et al. 2016Extra Resource=None, Type=maximum, pre-trained word embedding=true2017.09 | 95.93 | — | — | — | |
| JMT_ABtraining_mode=joint training2016.11 | 95.77 | — | — | — | |
| Hashimoto et al. 2016Extra Resource=PTB-POS, Type=reported2017.09 | 95.77 | — | — | — | |
| Hashimoto et al.2017.11 | 95.77 | — | — | — | |
| Søgaard and Goldberg2016.11 | 95.56 | — | — | — | |
| Søgaard et al. 2016Extra Resource=PTB-POS, Type=reported2017.09 | 95.56 | — | — | — | |
| Søgaard and Goldberg2017.11 | 95.56 | — | — | — | |
| Yang et al. 2017Extra Resource=CONLL 2000/PTB-POS dataset, Type=reported, pre-trained word embedding=true2017.09 | 95.41 | — | — | — | |
| LSTM-CRF+char-IntNet-5Character Embedding Model=IntNet-52018.10 | 95.29 | — | — | — | |
| Søgaard et al. 2016Extra Resource=1B Word dataset, Type=reported2017.09 | 95.28 | — | — | — | |
| Søgaard and Goldberg2019.06 | 95.28 | — | — | — | |
| BiLSTM-CRFAdversarial Training=true2017.11 | 95.25 | — | — | — | |
| Specialized HMM + voting scheme2015.08 | 95.23 | — | — | — | |
| BiLSTM-CRFAdversarial Training=false2017.11 | 95.18 | — | — | — | |
| Suzuki and Isozaki2016.11 | 95.15 | — | — | — | |
| Suzuki and Isozaki2017.11 | 95.15 | — | — | — | |
| LSTM-CRF+char-IntNet-9Character Embedding Model=IntNet-92018.10 | 95.08 | — | — | — | |
| CCNN+WLSTM+CRFCharacter sequence representation=CNN, Word sequence representation=LSTM, Inference layer=CRF2018.06 | 95.06 | — | — | — | |
| Singletraining_mode=single-task2016.11 | 95.02 | — | — | — | |
| Hashimoto et al. 2016Extra Resource=1B Word dataset, Type=reported2017.09 | 95.02 | — | — | — | |
| Hashimoto et al.2019.06 | 95.02 | — | — | — | |
| CLSTM+WLSTM+CRFCharacter sequence representation=LSTM, Word sequence representation=LSTM, Inference layer=CRF2018.06 | 95 | — | — | — | |
| Peters et al.Year=20172018.06 | 95 | — | — | — | |
| LSTM-CRF+char-LSTMCharacter Embedding Model=LSTM, Type=Re-implementation2018.10 | 94.97 | — | — | — | |
| LSTM-CRF+char-CNNCharacter Embedding Model=CNN, Type=Re-implementation2018.10 | 94.91 | — | — | — | |
| CCNN+WCNN+CRFCharacter sequence representation=CNN, Word sequence representation=CNN, Inference layer=CRF2018.06 | 94.77 | — | — | — | |
| CLSTM+WCNN+CRFCharacter sequence representation=LSTM, Word sequence representation=CNN, Inference layer=CRF2018.06 | 94.76 | — | — | — | |
| Zhai et al.2019.06 | 94.72 | — | — | — | |
| Yang et al.Year=20172018.06 | 94.66 | — | — | — | |
| Yang et al. 2017Extra Resource=1B Word dataset, Type=reported2017.09 | 94.66 | — | — | — | |
| Yang et al.2017.11 | 94.66 | — | — | — | |
| GRM-CRF+char-GRUCharacter Embedding Model=GRU, Reference=Yang et al., 20172018.10 | 94.66 | — | — | — | |
| Yang et al.2019.06 | 94.66 | — | — | — | |
| Nochar+WLSTM+CRFCharacter sequence representation=None, Word sequence representation=LSTM, Inference layer=CRF2018.06 | 94.49 | — | — | — | |
| Lample et al. 2016Extra Resource=None, Type=maximum, pre-trained word embedding=true2017.09 | 94.49 | — | — | — | |
| BI-LSTM-CRFEmbedding=Senna2015.08 | 94.46 | — | — | — | |
| BI-LSTM-CRF2018.05 | 94.46 | — | — | — | |
| LSTM-CRF+LexiconLexicon Usage=true, Reference=Huang et al., 20152018.10 | 94.46 | — | — | — | |
| Huang et al.external task-specific resources=true2019.06 | 94.46 | — | — | — | |
| Second order CRFReference=Sun et al., 20082015.08 | 94.34 | — | — | — | |
| Rei 2017Extra Resource=None, Type=max, pre-trained word embedding=true, leverage language models=true2017.09 | 94.33 | — | — | — | |
| Conv network taggerEmbedding=Senna2015.08 | 94.32 | — | — | — | |
| CNN-CRF2018.05 | 94.32 | — | — | — | |
| Collobert et al.2016.11 | 94.32 | — | — | — | |
| Collobert et al.2017.11 | 94.32 | — | — | — | |
| Conv-CRF+LexiconLexicon Usage=true, Reference=Collobert et al., 20112018.10 | 94.32 | — | — | — | |
| Collobert et al.external task-specific resources=true2019.06 | 94.32 | — | — | — | |
| Second order CRFReference=Sha and Pereira, 20032015.08 | 94.3 | — | — | — | |
| Second order CRFReference=Mcdonald et al., 20052015.08 | 94.29 | — | — | — | |
| LSTM-CRFType=Baseline2018.10 | 94.29 | — | — | — | |
| Nochar+WCNN+CRFCharacter sequence representation=None, Word sequence representation=CNN, Inference layer=CRF2018.06 | 94.23 | — | — | — | |
| BI-LSTM-CRF2015.08 | 94.13 | — | — | — | |
| SVM classifierYear=20012015.08 | 93.91 | — | — | — | |
| Kudo and Matsumoto2016.11 | 93.91 | — | — | — | |
| Lample et al. 2016Extra Resource=None, Type=reported, pre-trained word embedding=true2017.09 | 93.88 | — | — | — | |
| Rei2019.06 | 93.88 | — | — | — | |
| Tsuruoka et al.2016.11 | 93.81 | — | — | — | |
| Tsuruoka et al.2017.11 | 93.81 | — | — | — | |
| SVM classifierYear=20002015.08 | 93.48 | — | — | — | |
| CRFEncoder=WORD-CHAR-BILSTM2020.09 | 92.88 | — | — | — | |
| AIN-1OEncoder=WORD-CHAR-BILSTM2020.09 | 92.87 | — | — | — | |
| AIN-F2OEncoder=WORD-CHAR-BILSTM2020.09 | 92.85 | — | — | — | |
| MaxEntEncoder=WORD-CHAR-BILSTM2020.09 | 92.58 | — | — | — | |
| Context-to-Vecbackbone=BERT2022.10 | 91.98 | — | — | 92.8 | |
| SynGCN2022.10 | 91.23 | — | — | 92.23 | |
| BERT+Skip-gram2022.10 | 91.06 | — | — | 92.3 | |
| BERTavgpooling=average2022.10 | 90.96 | — | — | 90.76 | |
| SWEM-CRFcontextual_info=none before CRF2018.05 | 90.34 | — | — | — | |
| Conv-CRF2015.08 | 90.33 | — | — | — | |
| GloVe2022.10 | 89.87 | — | — | 91.84 | |
| CRFEncoder=WORD ONLY2020.09 | 89.39 | — | — | — | |
| CRFEncoder=WORD-CNN2020.09 | 89.21 | — | — | — | |
| AIN-1OEncoder=WORD-CNN2020.09 | 88.86 | — | — | — | |
| AIN-F2OEncoder=WORD ONLY2020.09 | 88.8 | — | — | — | |
| AIN-F2OEncoder=WORD-CNN2020.09 | 88.75 | — | — | — | |
| AIN-1OEncoder=WORD ONLY2020.09 | 88.69 | — | — | — | |
| Skip-gram2022.10 | 88.07 | — | — | 89.03 | |
| MaxEntEncoder=WORD-CNN2020.09 | 87.05 | — | — | — | |
| MaxEntEncoder=WORD ONLY2020.09 | 78.17 | — | — | — | |
| LD-NetLMs Ind.#=8, Pruning Status=pruned2018.04 | — | — | 13 | — | |
| LD-NetLMs Ind.#=9, Pruning Status=pruned2018.04 | — | — | 10 | — | |
| LD-Net *LMs Ind.#=8, Pruning=Trained without pruning (layer selection)2018.04 | — | 45.14 | 51 | — | |
| LD-Net *LMs Ind.#=9, Pruning=Trained without pruning (layer selection)2018.04 | — | 50.06 | 51 | — | |
| NoLMLMs Ind.#=None2018.04 | — | — | 3 | — | |
| R-ELMoLMs Ind.#=62018.04 | — | 40.27 | 108 | — | |
| R-ELMoLMs Ind.#=72018.04 | — | 48.85 | 68 | — |