Application of Text Mining Methods for Sentiment Analysis on Twitter Social Media

Abstract

Social media platforms, particularly Twitter, serve as significant channels for public expression, making sentiment analysis a crucial tool for understanding opinions, trends, and emotions in various domains. However, analyzing sentiment on Twitter presents challenges due to the short text length, informal language, slang, emojis, sarcasm, and multilingual content. This research applies text mining methods to enhance sentiment analysis on Twitter by developing improved text preprocessing techniques, hybrid machine learning models, and contextual sentiment classification approaches. The study explores various machine learning and deep learning models, including Naïve Bayes, Support Vector Machines (SVM), Long Short-Term Memory (LSTM), and Bidirectional Encoder Representations from Transformers (BERT), to evaluate their effectiveness in sentiment classification. A hybrid approach integrating machine learning and deep learning is proposed to enhance accuracy while maintaining computational efficiency. The research also introduces an adaptive text preprocessing method to normalize informal language, improve handling of code-mixed tweets, and enhance sarcasm detection using context-aware models with attention mechanisms. The results demonstrate that the hybrid classification model outperforms traditional machine learning and standalone deep learning models, achieving higher accuracy and better contextual understanding of sentiments. Furthermore, multilingual sentiment analysis and sarcasm detection improvements contribute to more reliable classification outcomes. The findings have practical implications for various sectors, including brand reputation analysis, political sentiment tracking, financial market predictions, and public health monitoring. This study contributes to the advancement of natural language processing (NLP), sentiment analysis, and social media analytics by proposing novel methodologies for handling Twitter-specific linguistic challenges. Future research directions include real-time sentiment analysis integration and expanding multimodal sentiment detection using text, images, and videos.