Joint Credibility Estimation of News, User, and Publisher via Role-Relational Graph Convolutional Networks

This repository contains the code for the paper "Joint Credibility Estimation of News, User, and Publisher via Role-Relational Graph Convolutional Networks"

Data Sources

FakeNewsNet-PolitiFact dataset can be downloaded using the code provided at https://github.com/KaiDMML/FakeNewsNet

PolitiFact-2021 dataset can be downloaded using the instructions provided at Data/README.md

Content

• Script to generate train-test split : Code/Utils/five_fold_train_test_split.ipynb
• Script to run baseline models for FakeNewsNet-PolitiFact dataset: Code/Experiments/FakeNewsNet-PolitiFact/Baseline_exp.ipynb
• Script to run Role-RGCN model for FakeNewsNet-PolitiFact dataset: Code/Experiments/FakeNewsNet-PolitiFact/Role_RGCN_exp.ipynb
• • Script to run baseline models for PolitiFact-2021 dataset: Code/Experiments/PolitiFact-2021/Baseline_exp.ipynb
• Script to run Role-RGCN model for PolitiFact-2021 dataset: Code/Experiments/PolitiFact-2021/Role_RGCN_exp.ipynb
• Helper files: Utils/features.py, Models/models.py