Addressing Vaccine Misinformation on Social Media by leveraging Transformers and User Association Dynamics
Published in ICMLDE, 2023
Implemented and trained BERT & XLNet to classify misinformation on custom datasets scraped from Twitter and Reddit. Fine tuned the architectures to obtain a F1-Score of 0.92 for BERT & 0.93 for XLNet. Implemented a user association mapping algorithm based on cosine similarity to identify dense conversation clusters, analyze conversations and flag suspicious users. The codebase is made publicly available on GitHub. The paper can be downloaded Here.
Recommended citation: Chirag Rao, Gautham Manuru Prabhu, Ajay Rajendra Kumar, Shourya Gupta and Nisha P. Shetty