Papers with news documents
MMUTF: Multimodal Multimedia Event Argument Extraction with Unified Template Filling (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent MEE methods focus on weak alignment strategies and data augmentation with simple classification models. |
| Approach: | They propose a unified template filling model that connects textual and visual modalities via textual prompts. |
| Outcome: | The proposed model surpasses the current SOTA on textual EAE by +7% F1 and performs generally better than the second-best systems for multimedia EAE. |