In Journal of ambient intelligence and humanized computing
The latest buzzword in today's world is fake news. The circulation of false information influences elections, public health, brand reputations, and violence. Hence, the severity of the threat of fake news is increasing. The danger for fake news exists everywhere globally and is not specific to one language or nation. The creators of fake news layer the facts in the news with misinformation to confuse the readers. Hence, a need arises for creating a model for detecting fake news in multiple languages. This paper proposes a unified attention-based model Mul-FaD to detect fake news in various languages. We have created our dataset with around 40000 articles in English, German, and French. This paper also shows an exploratory analysis of the dataset created. In this paper, we perform experiments from a multilingual perspective in which we use an altered hierarchical attention-based network to detect fake news. Our model is able to achieve an accuracy of 93.73 and an F1 score of 92.9 for the combined corpus of the three languages.
Ahuja Nishtha, Kumar Shailender
2023-Jan-14
Deep learning, Fake news, Natural language processing, Social media
