Summarization on CNN/DailyMail human evaluation (100 samples)
43Relevance ScoreSoft Layer-Specific Multi-Task Summarization (MTL)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Soft Layer-Specific Multi-Task Summarization (MTL)Comparison Pair=MTL vs. BASELINE, Evaluation Protocol=Pairwise Human Evaluation (wins count)2018.05 | 43 | 40 | 83 | |
| Soft Layer-Specific Multi-Task Summarization (MTL)Comparison Pair=MTL VS. SEE ET AL. (2017), Evaluation Protocol=Pairwise Human Evaluation (wins count)2018.05 | 39 | 33 | 72 | |
| Non-distinguishableComparison Pair=MTL vs. BASELINE, Evaluation Protocol=Pairwise Human Evaluation (count)2018.05 | 35 | 36 | 71 | |
| Non-distinguishableComparison Pair=MTL VS. SEE ET AL. (2017), Evaluation Protocol=Pairwise Human Evaluation (count)2018.05 | 32 | 29 | 61 | |
| Pointer-Coverage (See et al., 2017)Comparison Pair=MTL VS. SEE ET AL. (2017), Evaluation Protocol=Pairwise Human Evaluation (wins count)2018.05 | 29 | 38 | 67 | |
| BaselineComparison Pair=MTL vs. BASELINE, Evaluation Protocol=Pairwise Human Evaluation (wins count)2018.05 | 22 | 24 | 46 |