Dr. MANIMANNAN G.

Sentiment Analysis, VADER, Social Media, Natural Language Processing, Text Classification

This study presents a systematic approach for analyzing sentiment in social media text using the Valence Aware Dictionary and sEntiment Reasoner (VADER) available through the Natural Language Toolkit (NLTK). The study uses a dataset containing 500 social media entries collected across the period from 2010 to 2023, with sentiments categorized as positive, negative, and neutral. The proposed methodology involves preprocessing the textual data, applying the VADER sentiment intensity analyzer, calculating positive, negative, neutral, and compound sentiment scores, and assigning sentiment labels based on compound-score thresholds. The predicted sentiments are compared with the actual sentiment categories using cross-tabulation and graphical visualizations. The results indicate that positive sentiment is predominant, with 242 instances, followed by neutral sentiment with 144 instances and negative sentiment with 113 instances. The comparative analysis shows that VADER performs particularly well in identifying positive sentiments, correctly classifying 147 out of 166 actual positive instances, while greater difficulty is observed in distinguishing negative and neutral expressions. The study demonstrates the usefulness of VADER for examining emotional patterns in short social media texts and provides visual and numerical representations that support interpretation of sentiment trends.