Fake-News-Detection

CSE 573: Semantic Web Mining Project

Group 10

• Abhijith Shreesh (ASU ID: 1213204276)
• Aditya Chayapathy (ASU ID: 1213050538)
• Anuhya Sai (ASU ID: 1212931887)
• Arun Karthick Manickam Alagar Muthumanickam (ASU ID: 1213135077)
• Jagdeesh Basavaraju (ASU ID: 1213004713)

Description

The project aims at classifying the given news articles as fake or true based on the content and users associated with it using Graph Attention Networks (GATs).

• Extracted the content of news articles from the given dataset.
• Vectorized the news article content using BERT to obtain feature vector for every article.
• Derived relationship among news articles based on the users the articles are associated with.
• Classified the news articles by feeding the feature vectors and relationship matrix to the GAT.
• Compared and contrasted the performance of GAT against traditional machine learning algorithms.

Technology used: Google BERT, Graph Attention Network (GAT), Python, Pandas, NumPy, scikit-learn, Tensorflow

Steps to execute

• Go to the folder named "codebase".
• Run the command "pip install -r requirements.txt && python executebfpf.py BuzzFeed".
• The above command will install all the requirements and run GAT on Buzzfeed dataset.
• Run the command "python executebfpf.py PolitiFact".
• The above command will run GAT on PolitiFact dataset.
• After running the above commands on each dataset, results on training, validation and test set will be displayed.