Automated Misinformation Detection in Online News Using NLP and LSTM-Based Deep Learning

Abstract: The exponential growth of internet-based communication channels has revolutionized information accessibility, yet it has simultaneously intensified the circulation of deceptive, manipulated, and non-credible news content across digital platforms. The uncontrolled diffusion of misinformation can influence public perception, distort democratic processes, trigger financial instability, and create widespread social confusion, making automated verification systems increasingly essential in the modern data ecosystem. This project, titled Automated Misinformation Detection in Online News Using NLP and LSTM-Based Deep Learning, proposes an intelligent text analytics framework that leverages Natural Language Processing and deep sequential learning to distinguish authentic news articles from fabricated narratives. The system begins with structured dataset acquisition followed by extensive preprocessing operations such as text normalization, noise elimination, token segmentation, stop-word filtering, vocabulary encoding, and sequence standardization to transform raw textual inputs into computationally meaningful representations. An embedding layer is employed to learn semantic associations between words, after which a Bidirectional Long Short-Term Memory network captures contextual dependencies from both preceding and succeeding directions within a sentence, enabling superior understanding of narrative flow, linguistic irregularities, and deceptive writing patterns. The model is trained using labeled news corpora and evaluated through rigorous performance indicators including accuracy, precision, recall, F1-score, and confusion matrix interpretation, where the experimental results demonstrate strong predictive capability and robust generalization on previously unseen samples. Compared with conventional machine learning classifiers, the proposed architecture offers enhanced contextual comprehension and higher classification reliability for complex textual data. The modular design of the framework also supports practical deployment in news authentication portals, content moderation systems, browser-based verification tools, and large-scale media monitoring environments. Furthermore, the solution can be extended through multilingual processing, transformer integration, explainable artificial intelligence mechanisms, and real-time misinformation surveillance. Overall, the project establishes that the fusion of advanced NLP techniques with LSTM-driven deep learning forms a scalable, accurate, and future-ready approach for strengthening trust and credibility within digital information networks.

Keywords: Deep Learning, Detection, Accuracy and F-1 Score.