Abstractive Question Summarization on MATINF English translated subset (test)
48.17ROUGE-1ProphetNet + ROUGE-L
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ProphetNet + ROUGE-Lcategory=Baseline, RL_reward=ROUGE-L2021.07 | 48.17 | 28.13 | 48.66 | |
| ProphetNet + QTR + QFRcategory=Proposed Approach, reward=QTR + QFR2021.07 | 47.73 | 28.54 | 49.33 | |
| ProphetNet + QFRcategory=Proposed Approach, reward=Question-focus recognition2021.07 | 47.53 | 28.29 | 49.11 | |
| ProphetNet + QTRcategory=Proposed Approach, reward=Question-type identification2021.07 | 47.51 | 28.4 | 48.94 | |
| ProphetNet + Q-typecategory=Joint Learning2021.07 | 47.19 | 28.02 | 48.7 | |
| ProphetNet + Q-type + Q-focuscategory=Joint Learning2021.07 | 47.18 | 28.04 | 48.65 | |
| ProphetNet + Q-focuscategory=Joint Learning2021.07 | 47.14 | 28.06 | 48.64 | |
| ProphetNetcategory=Baseline2021.07 | 46.94 | 27.77 | 48.43 | |
| BART-LARGEcategory=Baseline2021.07 | 42.52 | 23.13 | 43.98 | |
| PEGASUScategory=Baseline2021.07 | 40.05 | 23.67 | 43.3 | |
| T5-BASEcategory=Baseline2021.07 | 39.66 | 21.24 | 41.52 | |
| MINILMcategory=Baseline2021.07 | 35.6 | 18.08 | 38.7 | |
| BertSummcategory=Baseline2021.07 | 31.16 | 11.94 | 34.7 | |
| SOTA*category=Baseline, implementation=trained on same data2021.07 | 24.58 | 7.3 | 28.08 | |
| Pointer Generator (PG)category=Baseline2021.07 | 23.31 | 7.01 | 26.61 | |
| Transformercategory=Baseline2021.07 | 22.25 | 5.89 | 26.06 | |
| Seq2Seq + Attentioncategory=Baseline2021.07 | 19.45 | 6.45 | 23.77 | |
| Seq2Seqcategory=Baseline2021.07 | 17.77 | 5.1 | 21.48 |