A Hybrid Deep Learning Framework using Distil RoBERTa for Sentiment Analysis of Russia–Ukraine War Discourse on Twitter

Abstract

With over 500 million tweets posted daily, a significant portion reflects opinions on global sociopolitical events such as the Russia–Ukraine War. Monitoring public sentiment at this scale through manual analysis is time-consuming, inconsistent, and impractical for real-time decision-making. To address this challenge, this study proposes a hybrid deep learning framework for automated sentiment analysis of Twitter discourse related to the Russia–Ukraine conflict. The methodology begins with comprehensive Natural Language Processing (NLP) preprocessing, including tokenization, stopword removal, normalization, and lemmatization, followed by Exploratory Data Analysis (EDA) to examine sentiment distribution, word frequencies, and textual trends. Context-aware semantic features are extracted using Distilled Robustly Optimized BERT Pretraining Approach (DistilRoBERTa), which provides efficient and lightweight transformer-based embeddings while reducing computational overhead. To mitigate class imbalance and improve minority class detection, KMeans-Synthetic Minority Over Sampling Technique (SMOTE) is employed for synthetic oversampling. For performance benchmarking, conventional classifiers such as Logistic Regression (LR), Support Vector Machine Classifier (SVC), and Random Forest Classifier (RFC) are implemented. The proposed model integrates a Deep Neural Network (DNN) for discriminative feature learning with a Greedy Tree-based classifier (GTC) to enhance classification accuracy and generalization. Hereafter, the proposed system named as “SemanticDistilDeepTree”. Experimental results demonstrate that the hybrid approach consistently outperforms baseline models across accuracy and F1-score metrics. The framework offers a scalable, efficient, and reliable solution for real-time sentiment monitoring, supporting policymakers, researchers, and analysts in understanding public perceptions of global conflicts