Summarization on CNN/Daily Mail Original and Anonymized (test)
40.34ROUGE-1lead-3 baseline
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| lead-3 baselineModel Type=Extractive, Dataset Version=Original2017.04 | 40.34 | 17.7 | 36.57 | 20.48 | 22.21 | |
| extractive modelModel Type=Extractive, Dataset Version=Anonymized2017.04 | 39.6 | 16.2 | 35.3 | — | — | |
| pointer-generator + coverageModel Type=Abstractive, Dataset Version=Original2017.04 | 39.53 | 17.28 | 36.38 | 17.32 | 18.72 | |
| lead-3 baselineModel Type=Extractive, Dataset Version=Anonymized2017.04 | 39.2 | 15.7 | 35.5 | — | — | |
| pointer-generatorModel Type=Abstractive, Dataset Version=Original2017.04 | 36.44 | 15.66 | 33.42 | 15.35 | 16.65 | |
| abstractive modelModel Type=Abstractive, Dataset Version=Anonymized2017.04 | 35.46 | 13.3 | 32.65 | — | — | |
| seq-to-seq + attn baselineModel Type=Abstractive, Vocabulary Size=50k, Dataset Version=Original2017.04 | 31.33 | 11.81 | 28.83 | 12.03 | 13.2 | |
| seq-to-seq + attn baselineModel Type=Abstractive, Vocabulary Size=150k, Dataset Version=Original2017.04 | 30.49 | 11.17 | 28.08 | 11.65 | 12.86 |