Abstractive Summarization on NYT ConvoSumm 1.0 (test)
36.6ROUGE-1BART-arg-filtered
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BART-arg-filteredInput representation=arg-filtered, Training points=2002021.06 | 36.6 | 9.83 | 32.61 | |
| BART-argPre-trained Model=BART-large, Training Samples=200 points, Input Format=Argument-mining input2021.06 | 36.6 | 9.83 | 32.61 | |
| BART-arg-graphInput representation=arg-graph, Training points=2002021.06 | 36.02 | 9.6 | 32.34 | |
| BARTInput representation=vanilla, Training points=2002021.06 | 35.91 | 9.22 | 31.28 | |
| BARTPre-trained Model=BART-large, Training Samples=200 points, Input Format=Vanilla2021.06 | 35.91 | 9.22 | 31.28 |