Multimodal Abstractive Text Summarization on How2 300h (test)
48.85ROUGE-1MAST
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MASTModalities=Video, Text, Audio2020.10 | 48.85 | 29.51 | 43.23 | 35.4 | |
| Video-TextModalities=Video, Text2020.10 | 48.4 | 27.97 | 42.23 | 32.89 | |
| TrimodalH2Modalities=Video, Text, Audio2020.10 | 47.85 | 28.46 | 42.17 | 35.65 | |
| MAST-BinnedModalities=Video, Text, Audio, Audio Feature Processing=Binned (average of 30 timesteps)2020.10 | 46.22 | 25.94 | 40.34 | 33.56 | |
| Text OnlyModalities=Text2020.10 | 46.01 | 25.16 | 39.98 | 33.45 | |
| Video OnlyModalities=Video2020.10 | 39.23 | 19.82 | 34.17 | 27.06 | |
| Audio-TextModalities=Audio, Text2020.10 | 34.56 | 15.22 | 31.63 | 28.36 | |
| BertSumAbsModalities=Text2020.10 | 29.68 | 11.74 | 22.58 | 31.53 | |
| Audio OnlyModalities=Audio2020.10 | 29.16 | 12.36 | 28.86 | 26.65 |