SENTIMENT ANALYSIS IN SOCIAL MEDIA: HOW DATA SCIENCE IMPACTS PUBLIC OPINION KNOWLEDGE INTEGRATES NATURAL LANGUAGE PROCESSING (NLP) WITH ARTIFICIAL INTELLIGENCE (AI)

This systematic literature review investigates the advancements, methodologies, challenges, and application domains of sentiment analysis with a particular focus on informal digital text such as social media content. A total of 91 peer-reviewed articles published between 2010 and 2024 were carefully selected and analyzed using the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) framework to ensure methodological rigor, transparency, and reproducibility. The review spans traditional machine learning algorithms, deep learning models, and transformer-based architectures, examining their role in enhancing sentiment classification accuracy across various textual and multimodal inputs. Key themes emerging from the analysis include the evolution of multimodal sentiment analysis incorporating emojis, images, and videos; the growing focus on emotion classification beyond polarity detection; and the development of multilingual and cross-lingual sentiment systems that aim to extend sentiment mining beyond English-dominated datasets. Furthermore, a notable subset of studies addressed the complexities of detecting sarcasm, irony, and linguistic ambiguity, highlighting significant limitations in even the most advanced models. The review also discusses the growing body of research in financial, political, and health-related sentiment analysis, where domain-specific customization has proven critical for reliable prediction. Despite technical progress, challenges remain in areas such as data imbalance, inconsistent evaluation metrics, lack of cross-domain generalizability, and insufficient attention to ethical concerns, including algorithmic bias and explainability. This review contributes a synthesized and critical understanding of the current state of sentiment analysis and identifies key research gaps, offering a valuable reference point for scholars, developers, and practitioners aiming to improve the robustness, inclusivity, and ethical grounding of sentiment analysis systems.