Publication

Abstract : Categorizing social media text data with high accuracy is an imposing challenge due to syntactic similarity but semantic inequality among them. This paper introduces a novel approach for grouping social media content data, capable of effectively managing both short and long texts. Short content is enriched with additional social media data and reliable news sources before being classified into specific categories, alongside long texts. The text is then categorized into five distinct categories, including government and governance, business and economy, crime, disaster, and entertainment. After the process of classification, cluster the categories to get distinct events. In comparison with traditional algorithms, this stratified approach improves cluster accuracy, yielding an impressive 99.1% accuracy for text categorization. After categorization, employing the silhouette score for clustering achieved a score close to 1 for each cluster. The approach leverages a large language model-based algorithm.