Abstractive Summarization on CNN (test)
31.9ROUGE-1BASE+E2Tcnn+sd
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BASE+E2Tcnn+sdbackbone=CNN, selective disambiguation=true2018.06 | 31.9 | 10.1 | 23.9 | |
| GBAtype=abstractive, architecture=graph-based attentional neural model2018.06 | 30.3 | 9.8 | 20 | |
| BASE+E2Trnn+sdbackbone=RNN, selective disambiguation=true2018.06 | 27.6 | 7.9 | 21.5 | |
| Distraction-M3type=abstractive, architecture=distraction-based network2018.06 | 27.1 | 8.2 | 18.7 | |
| BASE+E2Tcnnbackbone=CNN2018.06 | 26.6 | 7.3 | 20.7 | |
| BASE+E2Tcnn+softbackbone=CNN, attention=soft2018.06 | 26.6 | 7 | 20.6 | |
| BASE+E2Trnnbackbone=RNN2018.06 | 26.1 | 6.9 | 20.1 | |
| Lead-3type=extractive2018.06 | 26.1 | 9.6 | 17.8 | |
| LexRanktype=extractive2018.06 | 26.1 | 9.6 | 17.7 | |
| BASE: s2s+attarchitecture=seq2seq with attention2018.06 | 25.5 | 5.8 | 20 | |
| BASE+E2Trnn+softbackbone=RNN, attention=soft2018.06 | 25 | 6.7 | 19.8 | |
| Bi-GRUtype=abstractive, architecture=one-layer seq2seq2018.06 | 19.5 | 5.2 | 15 |