Multimedia Automatic Misogyny Identification for Task 5 of SemEval-2022, the 16h International Workshop on Semantic Evaluation
This project tackles Multimedia Automatic Misogyny Identification (MAMI), as outlined for Task 5 of SemEval-2022, the16th International Workshop on Semantic Evaluation. It addresses a popular topic of interest in the machine learning and natural language processing research communities -- multimodal inputs, in this case working with both images and text to produce meaningful outputs. The task is to identify misogynistic memes from a dataset that is provided by the challenge organizers.
- The caption generator files are in
Caption Generator
folder. - Overlaid Text removal files are in
Overlaid Text Remover
folder. - The BERT implementation files for text features (both overlaid text and captions) are in
Text Embedding
folder. - The visual embedding generator files are in
Visual Embedding
folder. - The datasets are not made public here because of license issues.
Note: Work still in progress