Dependency Parsing on Chinese Treebank (CTB) (test)
94.6UASMrini et al.
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Mrini et al.extra constituency annotation=true2023.06 | 94.6 | 89.3 | |
| LAL-Parser2019.11 | 94.56 | 89.28 | |
| HPSG+LAAdditional Data=Constituency Parsing2021.05 | 94.56 | 89.28 | |
| Yang and Ture-implementation using same pretrained encoder=true2023.06 | 93.5 | 92.5 | |
| Joint Span HPSGstrategy=Joint, data_split=constituent2019.07 | 93.24 | 91.95 | |
| Hexataggerbackbone=XLNet2023.06 | 93.2 | 91.9 | |
| Ours-ProjBackbone=BERT, Decoding Strategy=Projective2021.05 | 93.14 | 91.27 | |
| Ours-NprojBackbone=BERT, Decoding Strategy=Non-projective2021.05 | 93.06 | 91.21 | |
| SynTr+RNGTrBackbone=BERT2021.05 | 92.98 | 91.18 | |
| HiePTRBackbone=BERT2021.05 | 92.7 | 91.5 | |
| Ours-ProjBackbone=RoBERTa, Decoding Strategy=Projective2021.05 | 92.68 | 90.91 | |
| MulPTR+BERTAdditional Data=Constituency Parsing, Backbone=BERT2021.05 | 92.58 | 91.42 | |
| BiaffineBackbone=BERT2021.05 | 92.58 | 90.7 | |
| Ours-NprojBackbone=RoBERTa, Decoding Strategy=Non-projective2021.05 | 92.58 | 90.83 | |
| MP2OBackbone=BERT2021.05 | 92.55 | 90.69 | |
| BiaffineBackbone=RoBERTa2021.05 | 92.45 | 90.48 | |
| MP2OBackbone=RoBERTa2021.05 | 92.37 | 90.4 | |
| Division Span HPSGstrategy=Division, data_split=constituent2019.07 | 91.69 | 90.54 | |
| Joint Span HPSGstrategy=Joint2019.07 | 91.21 | 89.15 | |
| Zhou and Zhaoyear=2019, backbone=BERT2019.11 | 91.21 | 89.15 | |
| Zhou and Zhaoextra constituency annotation=true2023.06 | 91.2 | 89.2 | |
| Ours-ProjBackbone=None, Decoding Strategy=Projective2021.05 | 91.15 | 89.68 | |
| Ours-NprojBackbone=None, Decoding Strategy=Non-projective2021.05 | 91.12 | 89.55 | |
| GNNBackbone=None2021.05 | 90.78 | 89.5 | |
| HiePTRBackbone=None2021.05 | 90.76 | 89.67 | |
| MulPTRAdditional Data=Constituency Parsing2021.05 | 90.61 | 89.51 | |
| STACKPTRModel Type=Transition-based, Variant=Full2018.05 | 90.59 | 89.29 | |
| Ma et al.2019.07 | 90.59 | 89.29 | |
| Ma et al.year=20182019.11 | 90.59 | 89.29 | |
| StackPTRBackbone=None2021.05 | 90.59 | 89.29 | |
| STACKPTRModel Type=Transition-based, Variant=+gpar2018.05 | 90.49 | 89.19 | |
| STACKPTRModel Type=Transition-based, Variant=Org2018.05 | 90.48 | 89.19 | |
| BiAFModel Type=Graph-based, Variant=re-impl2018.05 | 90.43 | 89.14 | |
| STACKPTRModel Type=Transition-based, Variant=+sib2018.05 | 90.43 | 89.15 | |
| Biaffine+GlyceEmbeddings=Glyce-word, Backbone=Deep Biaffine model2019.01 | 90.2 | 89 | |
| Ensemble MST ParserParsing Approach=Transition (greedy), Pretrained word embeddings=true, Ensemble Size=202016.09 | 89.8 | 88.56 | |
| In-order parserSupervised reranking=true2017.07 | 89.4 | 88.4 | |
| BiAFModel Type=Graph-based, Variant=Original2018.05 | 89.3 | 88.23 | |
| Dozat and Manningyear=2017, Graph-based parsing=true2017.07 | 89.3 | 88.2 | |
| Dozat and Manning2019.07 | 89.3 | 88.23 | |
| BiaffineBackbone=Deep Biaffine model2019.01 | 89.3 | 88.2 | |
| Dozat and Manningyear=20162019.11 | 89.3 | 88.23 | |
| BiaffineBackbone=None2021.05 | 89.3 | 88.23 | |
| Dozat and Manning2023.06 | 89.3 | 88.2 | |
| Division Span HPSGstrategy=Division2019.07 | 89.14 | 87.31 | |
| Ma and HovyModel Type=Graph-based2018.05 | 89.05 | 87.74 | |
| Ma and Hovy2019.07 | 89.05 | 87.74 | |
| Ma and Hovyyear=20172019.11 | 89.05 | 87.74 | |
| Kuncoro et al.Model Type=Graph-based2018.05 | 88.87 | 87.3 | |
| Neural FOG, CD (distilled)Parsing Approach=Graph (1st order), Pretrained word embeddings=true, Distilled=true2016.09 | 88.87 | 87.3 | |
| Kuncoro et al.2019.07 | 88.87 | 87.3 | |
| Kuncoro et al.year=20162019.11 | 88.87 | 87.3 | |
| Li et al.2019.07 | 88.78 | 86.23 | |
| Li et al.year=20182019.11 | 88.78 | 86.23 | |
| Cheng et al.Model Type=Graph-based2018.05 | 88.1 | 85.7 | |
| Cheng et al.year=2016, Graph-based parsing=true2017.07 | 88.1 | 85.7 | |
| Cheng et al.2019.07 | 88.1 | 85.7 | |
| Cheng et al.2019.01 | 88.1 | 85.7 | |
| Ma and Zhao (2012)Parsing Approach=Graph (4th order), Pretrained word embeddings=false2016.09 | 87.74 | — | |
| Ballesteros et al.year=20162017.07 | 87.7 | 86.2 | |
| Ballesteros et al.2019.01 | 87.7 | 86.2 | |
| Ballesteros et al.Model Type=Transition-based, Year=20162018.05 | 87.65 | 86.21 | |
| Ballesteros et al. (2016)Parsing Approach=Transition (greedy), Pretrained word embeddings=true2016.09 | 87.65 | 86.21 | |
| Zhang et al.2019.07 | 87.65 | 86.17 | |
| BiLSTM Parsermethod=transition (greedy, dyn-oracle), representation=11 BiLSTM vectors, pretrained_embeddings=true2016.03 | 87.6 | 86.1 | |
| Ballesteros16method=transition (greedy, dyn-oracle), representation=Stack-LSTM + composition, pretrained_embeddings=true2016.03 | 87.6 | 86.2 | |
| Kiperwasser and GoldbergModel Type=Transition-based2018.05 | 87.6 | 86.1 | |
| Kiperwasser and Goldberg (2016)Parsing Approach=Transition (greedy), Pretrained word embeddings=true2016.09 | 87.6 | 86.1 | |
| Kiperwasser and Goldbergyear=20162017.07 | 87.6 | 86.1 | |
| Kiperwasser and Eliyahu2019.01 | 87.6 | 86.1 | |
| Wang and ChangModel Type=Graph-based2018.05 | 87.55 | 86.23 | |
| Wang and Chang (2016)Parsing Approach=Graph (1st order), Pretrained word embeddings=true2016.09 | 87.55 | 86.23 | |
| BiLSTM Parsermethod=transition (greedy, dyn-oracle), representation=4 BiLSTM vectors, pretrained_embeddings=true2016.03 | 87.4 | 85.9 | |
| In-order parserSupervised reranking=false2017.07 | 87.4 | 86.4 | |
| Neural FOG, CHParsing Approach=Graph (1st order), Pretrained word embeddings=true2016.09 | 87.32 | 85.82 | |
| Bohnet and NivreModel Type=Transition-based2018.05 | 87.3 | 85.9 | |
| Bohnet and Nivre (2012)Parsing Approach=Transition (beam), Pretrained word embeddings=false2016.09 | 87.3 | 85.9 | |
| Dyer et al.Model Type=Transition-based2018.05 | 87.2 | 85.7 | |
| Dyer et al. (2015)Parsing Approach=Transition (greedy), Pretrained word embeddings=true2016.09 | 87.2 | 85.7 | |
| Dyer15method=transition (greedy), representation=Stack-LSTM + composition, pretrained_embeddings=true2016.03 | 87.1 | 85.5 | |
| BiLSTM Parsermethod=graph, 1st order, representation=2 BiLSTM vectors, pretrained_embeddings=false2016.03 | 86.6 | 85.1 | |
| Kiperwasser and GoldbergModel Type=Graph-based2018.05 | 86.6 | 85.1 | |
| Kiperwasser and Goldberg (2016)Parsing Approach=Graph (1st order), Pretrained word embeddings=true2016.09 | 86.6 | 85.1 | |
| BiLSTM Parsermethod=transition (greedy, dyn-oracle), representation=11 BiLSTM vectors, pretrained_embeddings=false2016.03 | 86.5 | 84.9 | |
| BiLSTM Parsermethod=graph, 1st order, representation=2 BiLSTM vectors, pretrained_embeddings=true2016.03 | 86.5 | 84.9 | |
| BiLSTM Parsermethod=transition (greedy, dyn-oracle), representation=4 BiLSTM vectors, pretrained_embeddings=false2016.03 | 86.2 | 85 | |
| Ballesteros16method=transition (greedy, dyn-oracle), representation=Stack-LSTM + composition, pretrained_embeddings=false2016.03 | 86.1 | 84.5 | |
| ZhangNivre11method=transition (beam), representation=large feature set (sparse), pretrained_embeddings=false2016.03 | 86 | 84.4 | |
| Zhang and Nivre (2011)Parsing Approach=Transition (beam), Pretrained word embeddings=false2016.09 | 86 | 84.4 | |
| Dyer15method=transition (greedy), representation=Stack-LSTM + composition, pretrained_embeddings=false2016.03 | 85.7 | 84.1 | |
| Zhu15method=reranking /blend, representation=recursive conv-net, pretrained_embeddings=true2016.03 | 85.7 | — | |
| Zhu et al. (2015)Parsing Approach=Reranking/blend, Pretrained word embeddings=true2016.09 | 85.7 | — | |
| Dyer et al.year=20162017.07 | 85.5 | 84 | |
| Ballesteros et al.Model Type=Transition-based2018.05 | 85.3 | 83.72 | |
| Ballesteros et al. (2015)Parsing Approach=Transition (greedy), Pretrained word embeddings=false2016.09 | 85.3 | 83.72 | |
| Chen and ManningModel Type=Transition-based2018.05 | 83.9 | 82.4 | |
| Chen and Manning (2014)Parsing Approach=Transition (greedy), Pretrained word embeddings=true2016.09 | 83.9 | 82.4 | |
| Chen and Manning2019.07 | 83.9 | 82.4 | |
| Chen and Manning2023.06 | 83.9 | 82.4 |