AMR Parsing on LDC2015E86 (test)
70.7F1 Score5 bi-LSTM networks (word-based)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| 5 bi-LSTM networks (word-based)Model=5 bi-LSTM networks (word-based), Train set (gold)=LDC2015E862017.05 | 70.7 | — | — | |
| char-based seq2seq model + silverModel=char-based seq2seq model + silver, Train set (gold)=LDC2015E862017.05 | 68.5 | — | — | |
| SBMT2017.04 | 67.1 | — | — | |
| SBMTModel=SBMT, Train set (gold)=LDC2015E862017.05 | 67.1 | — | — | |
| JAMR (2016a)2017.04 | 67 | 69.7 | 64.5 | |
| JAMR-16Model=JAMR-16, Train set (gold)=LDC2015E862017.05 | 67 | — | — | |
| CAMR2017.04 | 66.5 | 70.4 | 63.1 | |
| CAMRModel=CAMR, Train set (gold)=LDC2015E862017.05 | 66.5 | — | — | |
| CCG*Note=Reported on LDC2014T12 newswire portion2017.04 | 66.3 | 66.8 | 65.7 | |
| AMR-eagerModel=AMR-eager, Train set (gold)=LDC2015E862017.05 | 64 | — | — | |
| Neural AMRSelf-training data=GIGA-20M2017.04 | 62.1 | 59.7 | 64.7 | |
| word-based seq2seq + gigaModel=word-based seq2seq + giga, Train set (gold)=LDC2015E862017.05 | 62.1 | — | — | |
| Neural AMRSelf-training data=GIGA-2M2017.04 | 61.9 | 60.2 | 63.6 | |
| Neural AMRSelf-training data=GIGA-200k2017.04 | 59.3 | 57.8 | 60.9 | |
| JAMR (2014)2017.04 | 58 | 64 | 53 | |
| Neural AMRTraining Data=AMR-ONLY2017.04 | 55.5 | 53.1 | 58.1 | |
| word-based seq2seqAuthors=Konstas et al. (2017), Model=word-based seq2seq, Train set (gold)=LDC2015E862017.05 | 55.5 | — | — | |
| SEQ2SEQ2017.04 | 52 | 55 | 50 | |
| word-based seq2seqAuthors=Peng et al. (2017), Model=word-based seq2seq, Train set (gold)=LDC2015E862017.05 | 52 | — | — | |
| CHAR-LSTM2017.04 | 43 | — | — | |
| char-based seq2seqModel=char-based seq2seq, Train set (gold)=LDC2015E862017.05 | 43 | — | — |