Fake News Detection using Deep Learning Models
- Post by: Bibek Upadhayay
- September 11, 2020
- Comments off
Over View
Social media has become an integral part of our daily lives and culture, providing unprecedented access to a constant flow of information. However, the unregulated nature of these platforms has made it easier for false information and fake news to spread rapidly. With the massive volume and speed of information on social media, it is practically impossible to manually monitor and control the spread of misinformation. To address this issue, we propose a novel approach using deep learning to automatically detect false short-text claims on social media.
Our approach, called Sentimental LIAR, expands the LIAR dataset of short claims by incorporating features based on sentiment and emotion analysis. We also introduce a new deep learning architecture using the BERT-Base language model to classify claims as genuine or fake. Our results indicate that our proposed architecture, when trained on Sentimental LIAR, achieves an impressive 70% accuracy – an improvement of about 30% over previously reported results for the LIAR benchmark. By using our approach, we can take a significant step towards mitigating the spread of false information and fake news on social media.
Current Team Members:
Dr. Vahid Behzadan,
Bibek Upadhayay
Tools and Dataset: We used HuggingFace Transformer DistilBERT and modified the LIAR Dataset to the Sentimental Liar and then performed the classification.
Publication: “Sentimental LIAR: Extended Corpus and Deep Learning Models for Fake Claim Classification” ARXIV