Ensemble-based fake news and disinformation detection using crowdsourced dataset

Fake content can manifest in two ways: intentional and unintentional. Intentional fake content is created with the intention of deceiving readers, while unintentional disinformation arises from the careless selection and dissemination of content. In result, the process of identifying fake information is complicated and requires a combination of different approaches. Firstly, it is essential to gather significant amounts of data from various sources, including social media, news outlets, and other online platforms. This data needs be further processed and analyzed using appropriate algorithms and tools to remove biases, noise, and invalid data samples. In this article, we present, first, a hybrid approach to feature extraction involving the use of DistilBERT and TF-IDF as an input vector for further analysis. Secondly, we developed a series of hybrid artificial intelligence models using a voting classifier to identify potentially fake news. An important advantage is the fact that in our approach we do not focus only on determining whether the message is true or false, but thanks to the developed dataset, we take into account many factors and determine the credibility of the content.