Synthesizing Political Context Understanding with AI Techniques to Improve Fake News Detection Performance across Regions

Abstract

Fake news is spreading quickly on the internet, which is very bad for society and the security of the government. The significant issue that was talked about in the paper was the creation of automatic systems that can detect fake news better and adapt to various areas. The dataset used in the study is the LIAR dataset, which is a standard set of various political statements labeled with varying degrees of truthfulness. Text is also cleaned up, tokenized, and represented with existing trained word embeddings such as GloVe and Word2Vec as a step in data preparation. To identify complex trends in the text, most language and contextual features are removed, such as syntactic, semantic, and sentiment-based ones. The primary contribution of this study is a way of grouping various features into one representation. A set of models is subjected to performance tests, and it includes Random Forest, Naive Bayes, Convolutional Neural Network (CNN), Autoencoder, and a proposed Hybrid CNN-Autoencoder architecture. The hybrid model performs the most, having the greatest precision and the most equalized classification scores. Comparative analysis demonstrates that the combination of deep learning and knowledge of the environment significantly enhances the level of detection in domains. It is a flexible AI-based system that can work in the context of language and political differences and is a big step forward in searching for fake information automatically.