GAME-ON

This is the source code for the paper "GAME-ON: Graph Attention Network based Multimodal Fusion for Fake News Detection."

Directory structure

1. creating-graph-data contains the code for creating node embeddings for textual and visual graphs.

2. unimodal-textual, unimodal-visual, and multimodal-concatenation folders contain code for ablation study.

3. game-on contains code for the proposed framework.

4. ModelSizeBaselines.ipynb is the code for roughly estimating the trainable paramters in the baselines.

Dataset

1. MediaEval 2015 (Twitter Dataset): Link to image-verificatiion corpus github

2. Weibo Dataset: Link to EANN paper for the dataset

Setup

Dependencies

1. dgl==0.7.0
2. torch==1.8.1
3. torchvision==0.9.1
4. transformers==4.10.2

Run the code

1. To run any of the available models, please move to the appropriate directory

2. run $ python main.py