Fake news detection using transformers
According to the Collins dictionary, fake news can be defined as 'false, often sensational, information disseminated under the guise of news reporting. Despite the fact that fake news existed for many years, its impact has recently increased. This trend can be easily observed, e.g., by means of Google Trends too. It shows that the phrase 'fake news' has rapidly become more popular since November 2016. Traditionally, fake news was known as rumors or propaganda, mostly used in order to make political or economic gains. The main goal of creating fake news has remained unchanged. However, currently it can spread more easily thanks to the popularity of social networks. The current pandemic reality has led to a serious outbreak of misinformation. It can be very dangerous in social, health-care and political aspects, like in the case of the fake news concerning the COVID-19 pandemic and its connection with the 5G transmission. It is your task to try to distinguish fake and real information!
Image source: https://www.analyticsvidhya.com/blog/2020/12/fake-news-classifier-on-us-election-news%F0%9F%93%B0-lstm-%F0%9F%88%9A/
Datasets source:
- WELFake https://www.kaggle.com/datasets/saurabhshahane/fake-news-classification
- Fake & Real https://www.kaggle.com/datasets/clmentbisaillon/fake-and-real-news-dataset?select=True.csv
- Fake news net https://www.kaggle.com/datasets/algord/fake-news
Used:
- Pandas profiling
- NLTK
- SentenceTransformer
- PyTorch
- StratifiedKFold
- Optuna