Hybrid Transformer-Based Semantic Feature Learning for Real-Time Political Event Detection from Twitter Streams

Abstract

The extensive use of Twitter as a real-time social media platform has led to the continuous generation of massive volumes of short-text messages related to political, social, and emergency events. Extracting meaningful insights from these rapidly evolving tweets is a challenging task because of their unstructured format, contextual ambiguity, linguistic diversity, and high velocity. Conventional machine learning techniques often fail to capture the rich semantic and contextual information embedded in short-text data, resulting in reduced event classification performance. To address these limitations, this study proposes an intelligent political event detection framework that integrates SBERTenhanced Lightweight RoBERTa with advanced machine learning algorithms. Initially, the collected tweets undergo comprehensive Natural Language Processing (NLP) preprocessing, including normalization, tokenization, stop-word removal, and lemmatization, to improve textual quality and consistency. Exploratory Data Analysis (EDA) is then performed to analyze data characteristics and identify meaningful linguistic patterns. A lightweight SBERTbased RoBERTa model subsequently generates semantic-rich contextual embeddings that serve as effective feature representations. These embeddings are utilized by multiple classifiers, including Stochastic Gradient Descent (SGD), Histogram Gradient Boosting (HGB), Random Forest Classifier (RFC), and Greedy Tree Classifier (GTC). Furthermore, a Deep Neural Network (DNN) performs intermediate feature learning, followed by an optimized SGD classifier to enhance predictive capability. The proposed framework classifies tweets into six event categories: disaster, political, positive, protest, riot, and terror. Experimental results demonstrate high classification accuracy, robustness, scalability, and strong generalization ability, highlighting the effectiveness of the proposed framework for real-time political event detection and intelligent decision-support applications.