CCG Supertagging on CCGBank (test)
96.29AccuracyHeterogeneous Dynamic Convolutions
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Heterogeneous Dynamic Convolutions2022.03 | 96.29 | 96.61 | 72.06 | 34.45 | 4.55 | |
| BERT + A-GCN (Chunk)Encoder=uncased BERT, Parser=EasyCCG2020.10 | 96.25 | — | — | — | — | |
| Attentive Convolutions2022.03 | 96.25 | 96.64 | 71.04 | — | — | |
| BERT Token Classification2022.03 | 96.22 | 96.58 | 70.29 | 23.17 | — | |
| Clark et al.2020.10 | 96.1 | — | — | — | — | |
| Cross-View Training2022.03 | 96.1 | — | — | — | — | |
| Recursive Tree Addressing2022.03 | 96.09 | 96.44 | 68.1 | 37.4 | 3.03 | |
| BERTEncoder=uncased BERT, Parser=EasyCCG2020.10 | 96.06 | — | — | — | — | |
| Symbol Sequential LSTM /w n-gram oracles2022.03 | 95.99 | 96.4 | 65.83 | 8.65 | — | |
| BERT + A-GCN (Full)Encoder=uncased BERT, Parser=EasyCCG2020.10 | 95.91 | — | — | — | — | |
| Stanojević and Steedman2020.10 | 95.4 | — | — | — | — | |
| Lewis et al.Architecture=BiLSTM-softmax, Training=Semi-supervised tri-training2019.08 | 94.7 | — | — | — | — | |
| BiLSTM-LANBackbone=BiLSTM, Decoder=LAN2019.08 | 94.7 | — | — | — | — | |
| Lewis et al. (2016)Semi-supervised tri-training=true2019.08 | 94.7 | — | — | — | — | |
| BiLSTM-LAN2019.08 | 94.7 | — | — | — | — | |
| Lewis et al.2020.10 | 94.7 | — | — | — | — | |
| Vaswani et al. (b)Components=LSTM language model and BiLSTM over supertags2019.08 | 94.5 | — | — | — | — | |
| Vaswani et al. (2016b)2019.08 | 94.5 | — | — | — | — | |
| Tu and GimpelBackbone=BiLSTM-CRF, Components=Inference network2019.08 | 94.4 | — | — | — | — | |
| Tu and Gimpel (2019)2019.08 | 94.4 | — | — | — | — | |
| Lewis et al.Architecture=BiLSTM-softmax2019.08 | 94.3 | — | — | — | — | |
| Lewis et al. (2016)Semi-supervised tri-training=false2019.08 | 94.3 | — | — | — | — | |
| Vaswani et al.2020.10 | 94.24 | — | — | — | — | |
| Vaswani et al. (a)Architecture=BiLSTM-softmax2019.08 | 94.2 | — | — | — | — | |
| Vaswani et al. (2016a)2019.08 | 94.2 | — | — | — | — | |
| BiLSTM-softmaxBackbone=BiLSTM, Decoder=softmax2019.08 | 94.1 | — | — | — | — | |
| BiLSTM-CRFBackbone=BiLSTM, Decoder=CRF2019.08 | 94.1 | — | — | — | — | |
| BiLSTM-softmax2019.08 | 94.1 | — | — | — | — | |
| BiLSTM-CRF2019.08 | 94.1 | — | — | — | — | |
| Søgaard and GoldbergArchitecture=BiRNN, Training=Multi-task learning2019.08 | 93.3 | — | — | — | — | |
| Søgaard and Goldberg (2016)2019.08 | 93.3 | — | — | — | — | |
| Xu et al.Architecture=BiRNN-softmax2019.08 | 93 | — | — | — | — | |
| Xu et al. (2015)2019.08 | 93 | — | — | — | — | |
| Xu et al.2020.10 | 93 | — | — | — | — | |
| Lewis and Steedman2020.10 | 91.3 | — | — | — | — |