Running summarizing tools on a new article

Check the paper below, then start its synthesis.


High-Dimensional Text Classification Using TF-IDF Features and Deep Neural Networks

Sudeep S S (KLE Technological University), Chetan S H (KLE Technological University), B. Gurubasavaraja (KLE Technological University), Nandeesh I B (KLE Technological University), Uday Kulkarni (KLE Technological University)

World Conference on Computational Science and Technology (WcCST)
License: not stated (all rights reserved)

Abstract: The rapid rise of digital text data in news articles, forums, and online communication has created the need to develop more accurate text classifying systems. Term-weighting techniques, such as Term Frequency-Inverse Document Frequency (TF-IDF), are the most commonly used text classifying techniques due to their simplicity and effectiveness; however, they are often faced with the challenge of high-dimensional noise, class imbalance, and class overlap in text data. To assess the effectiveness of the term weighting technique in modern text classifying systems, we propose the use of a regularized Multi-Layer Perceptron (MLP) classifier, which is trained using the TF-IDF feature representation of the widely used 20 Newsgroups dataset, which is considered the gold standard in text categorization tasks. Our proposed system is designed to reduce the risk of overfitting in high-dimensional feature spaces using L2 regularization, dropout, and early stopping techniques, which are essential in the performance of the proposed system. We show that the proposed system has high accuracy in text categorization tasks, achieving accuracy and macro F1 scores of 93% and 0.93, respectively, outperforming the Naive Bayes and Logistic Regression classifiers and achieving comparable performance with the Linear SVM classifier.

Published on 26 Mar. 2026