Sentiment` analysis, a key task in natural language processing (NLP), involves recognizing and classifying emotions expressed in textual data. By analyzing online discussions, companies can gain insights into public sentiment toward their offerings. With the rise of digital platforms like blogs and social media, people now frequently share reviews and opinions related to daily activities. This study presents a comparative analysis of sentiment analysis techniques by reviewing 15 research papers encompassing machine learning (ML), deep learning (DL), and traditional NLP approaches. The analysis highlights the evolving landscape of sentiment classification, showcasing the performance of models such as Support Vector Machines (SVM), Naïve Bayes, Random Forest (RF), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and advanced architectures like Capsule Networks and Switch Transformers. The studies were evaluated using common metrics such as precision, recall, accuracy, and F1-score. DL methods, particularly those combining LSTM and CNN with optimization strategies, performed better across all metrics, including precision, recall, accuracy, and F1-score. These hybrid DL models showed strong contextual understanding and handled complex data well. Traditional ML models offered interpretability and ease of implementation, while NLP-based methods remained competitive for lightweight applications. The review also notes challenges like sarcasm, domain shifts, and imbalanced data that hinder generalization. This comparative review provides insights into the strengths, limitations, and suitable application contexts of each approach, offering a foundation for selecting appropriate sentiment analysis methods in future research and real- world applications.
Buy article PDF