Comparative Analysis of Nlp Architectures For Emotion Recognition In Social Media: From Svm To Transformers

Social networks have evolved into massive repositories of emotional data, offering unique insights into human behavior. However, automating emo-tion recognition in these environments remains a challenge due to linguistic ambiguity, sarcasm, and the informality of digital text. This paper presents a comprehensive analysis of the state-of-the-art Natural Language Processing (NLP) algorithms used for this task, contrasting traditional Machine Learn-ing approaches with modern Deep Learning architectures. The analysis re-veals that while Support Vector Machines (SVM) remain relevant for spe-cific tasks, Transformer-based models such as BERT and RoBERTa have es-tablished a new standard of performance by effectively capturing semantic context. Furthermore, the study identifies critical limitations in current public datasets, particularly regarding cultural bias and the scarcity of re-sources for low-resource languages. The findings suggest that future research must prioritize multimodal approaches and the development of lightweight models for real-time edge computing.