Real Time Sentiment Analysis and Categorisation of News Articles Using VADER and SVM
S Indhumathi, F. Mary Harin Fernandez · 2025
This study introduces a novel framework for real-time sentiment analysis (SA) and categorization of news articles using a combination of VADER (Valence Aware Dictionary and sEntiment Reasoner) and Support Vector Machines (SVM). In a time when information spreads quickly, understanding the underlying sentiments in news content is crucial, particularly when the information spans diverse domains such as education, finance, and health. The proposed methodology connects real-time data acquired from APIs, ensuring that the analysis reflects current trends and public opinion dynamics. The first phase of the framework involves data collection from APIs specifically personalized to gather news articles across the three targeted domains. Once the data is obtained, a preprocessing pipeline cleans and normalizes the text, handling issues such as punctuation, stop words, and tokenization to ensure that the subsequent analysis is accurate and efficient. VADER is then applied to each article to perform sentiment scoring. VADER's rule-based approach, specifically tuned for social media and contemporary language, enables it to quickly assess the polarity of the text, identifying positive, negative, or neutral sentiments with high precision. After sentiment extraction, the framework employs the SVM algorithm to classify articles into one of the three categories: education, finance, or health. This classification utilizes a feature set that integrates both the sentiment scores from VADER and domain-specific textual features extracted from the articles. A comprehensive training process on a curated dataset, free from dependence on any preexisting published database, ensures that the SVM classifier is robust and tailored to the unique characteristics of the input data. Experimental results demonstrate that the integration of VADER and SVM facilitates a reliable and scalable solution for real-time sentiment analysis (SA) and domain categorization. The results underline the potential of this combined approach to support decision-makers in media monitoring and public relations, offering timely insights into the evolving landscape of public sentiment across critical sectors.