The rapid growth of digital content has made information retrieval and classification more challenging than ever. This paper presents a deep learning framework for news topic classification that integrates Natural Language Processing (NLP) techniques with machine learning models such as Logistic Regression, Random Forest, Ridge Classifier, and XGBoost, combined with an attention mechanism for improved feature selection. To address the issue of large data volumes, we propose a novel data reduction methodology that extracts key sentences from news articles, effectively reducing the data size and leading to a 50% reduction in inference time. Our framework maintains the accuracy of the classification model, achieving results comparable to existing approaches. Experimental evaluations using popular datasets like AG News demonstrate the superior performance of our system in terms of classification accuracy and efficiency, offering a scalable solution for real-time news classification.

Keywords: News Topic Classification, Natural Language Processing, Deep Learning, Logistic Regression, Random Forest, Ridge Classifier, XGBoost, Attention Mechanism, Data Volume Reduction, Inference Time Optimization, AG News Dataset, Model Efficiency.