SemEval-Task5-MAMI

Implementation of our SemEval-Task5-MAMI challenge paper. Different uni-modal and multi-modal models are trained. Finally, we use the ensemble learning strategy to combine the results at different levels.

How to run the code

• run run.sh from the beginning

File architecture

• processed dataset and features are saved in ./Dataset
• trained models are saved in ./model
• results are saved in ./results

Cite as

@misc{https://doi.org/10.48550/arxiv.2204.03953,
  url = {https://arxiv.org/abs/2204.03953},
  author = {Yu, Wentao and Boenninghoff, Benedikt and Roehrig, Jonas and Kolossa, Dorothea},  
  title = {RubCSG at SemEval-2022 Task 5: Ensemble learning for identifying misogynous MEMEs},
  publisher = {arXiv},
  year = {2022}
}