Abstractive Summarization on Gigaword full-length F1 (test)
37.04ROUGE-1 F1BASE+E2Tcnn+sd
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BASE+E2Tcnn+sdE2T Module=CNN, Selective Disambiguation=true2018.06 | 37.04 | 16.66 | 34.93 | |
| BASE+E2Trnn+sdE2T Module=RNN, Selective Disambiguation=true2018.06 | 36.89 | 16.86 | 34.74 | |
| BASE+E2TcnnE2T Module=CNN, Selective Disambiguation=false2018.06 | 36.56 | 16.56 | 34.57 | |
| BASE+E2Tcnn+softE2T Module=CNN, Soft Attention=true2018.06 | 36.56 | 16.44 | 34.58 | |
| BASE+E2TrnnE2T Module=RNN, Selective Disambiguation=false2018.06 | 36.52 | 16.21 | 34.32 | |
| BASE+E2Trnn+softE2T Module=RNN, Soft Attention=true2018.06 | 36.38 | 16.12 | 34.2 | |
| SEASSModel=BiGRU encoder and GRU decoder with selective encoding2018.06 | 36.15 | 17.54 | 33.63 | |
| BASE: s2s+attModel=Sequence-to-sequence with attention baseline2018.06 | 34.14 | 15.44 | 32.47 | |
| RAS-ElmanModel=Attentive CNN encoder with Elman RNN decoder2018.06 | 33.78 | 15.97 | 31.15 | |
| Luong-NMTModel=Two-layer LSTM encoder-decoder2018.06 | 33.1 | 14.45 | 30.71 | |
| Feat2sModel=RNN seq2seq with lexical/statistical features2018.06 | 32.67 | 15.59 | 30.64 | |
| ABS+Model=Fine-tuned version of ABS2018.06 | 29.78 | 11.89 | 26.97 |