An Entity-Aware Framework for Fake News Detection in Malayalam Using Machine Learning Techniques

Abstract

The rapid spread of misinformation on digital platforms has made fake news detection a critical challenge, particularly for low-resource languages such as Malayalam. Existing approaches rely mainly on textual and statistical features, which often fail to capture semantic inconsistencies in news content. This paper proposes an entity-aware framework that integrates machine learning techniques with transformer-based Named Entity Recognition (NER) to improve fake news detection. The framework extracts entity-level features such as entity frequency, relevance, and consistency, and combines them with textual representations like TF-IDF. These features are used to train machine learning classifiers including Support Vector Machine, Logistic Regression, and Random Forest. Experimental results demonstrate that the proposed entity-aware approach significantly improves classification performance compared to baseline models. The framework provides an efficient and scalable solution for misinformation detection in Malayalam.