Constituency Parsing on CTB7 (test)
91.73PrecisionNeural CRF Parser
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Neural CRF ParserEmbeddings=BERT2020.08 | 91.73 | 91.38 | 91.55 | — | — | — | |
| Neural CRF ParserPre-trained embeddings=true2020.08 | 88.84 | 88.36 | 88.6 | — | — | — | |
| Neural CRF ParserEmbeddings=Random2020.08 | 87.66 | 87.21 | 87.43 | — | — | — | |
| Bottom-up Pointer NetworksBackbone=BERT, Parsing Paradigm=Seq2Seq-based [Q]2021.10 | — | — | — | 91.66 | 91.31 | 91.49 | |
| Zhang et al. (2020)Backbone=BERT, Parsing Paradigm=Span-based [S]2021.10 | — | — | — | 91.73 | 91.38 | 91.55 |