Abstractive Summarization on Gigawords (test)
36.96ROUGE-1Risk
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RiskRisk optimization metric target=RG-1, Objective level=Sequence-level [S]2017.11 | 36.96 | 17.61 | 34.18 | |
| RiskRisk optimization metric target=RG-L, Objective level=Sequence-level [S]2017.11 | 36.7 | 17.88 | 34.29 | |
| RiskRisk optimization metric target=RG-2, Objective level=Sequence-level [S]2017.11 | 36.65 | 18.32 | 34.07 | |
| RNN MRTObjective level=Sequence-level [S]2017.11 | 36.54 | 16.59 | 33.44 | |
| TokLSObjective level=Token-level [T], description=Baseline with label smoothing2017.11 | 36.53 | 18.1 | 33.93 | |
| DRGD2018.05 | 36.3 | 17.6 | 33.6 | |
| seq2seq + CGUImplementation=Our implementation, Enhanced with=Convolutional Gated Unit2018.05 | 36.3 | 18 | 33.8 | |
| WFEObjective level=Token-level [T]2017.11 | 36.3 | 17.31 | 33.88 | |
| DRGD2017.08 | 36.27 | 17.57 | 33.62 | |
| DRGDObjective level=Token-level [T]2017.11 | 36.27 | 17.57 | 33.62 | |
| SEASS2018.05 | 36.2 | 17.5 | 33.6 | |
| SEASSObjective level=Token-level [T]2017.11 | 36.15 | 17.54 | 33.63 | |
| lvt5k-1sentVocabulary Size=5k, Input Sentences=1, Linguistic Features=true2017.08 | 35.3 | 16.64 | 32.62 | |
| ASC + FSC12017.08 | 34.17 | 15.94 | 31.92 | |
| RAS-Elman2018.05 | 33.8 | 16 | 31.2 | |
| RAS-ElmanArchitecture=Elman RNN2017.08 | 33.78 | 15.97 | 31.15 | |
| seq2seqImplementation=Our implementation2018.05 | 33.6 | 16.3 | 31.3 | |
| Feats2018.05 | 32.7 | 15.6 | 30.6 | |
| lvt2k-1sentVocabulary Size=2k, Input Sentences=1, Linguistic Features=true2017.08 | 32.67 | 15.59 | 30.64 | |
| RNN MLEObjective level=Token-level [T]2017.11 | 32.67 | 15.23 | 30.56 | |
| RAS-LSTM2018.05 | 32.6 | 14.7 | 30 | |
| RAS-LSTMArchitecture=LSTM2017.08 | 32.55 | 14.7 | 30.03 | |
| ABS+2018.05 | 29.8 | 11.9 | 27 | |
| ABS+2017.08 | 29.78 | 11.89 | 26.97 | |
| ABS+Objective level=Token-level [T]2017.11 | 29.78 | 11.89 | 26.97 | |
| ABS2018.05 | 29.6 | 11.3 | 26.4 | |
| ABS2017.08 | 29.55 | 11.32 | 26.42 |