Real-Time Sentiment Analytics: Integrating NLP for Social Media Insight
Senthil Pandi S, Pramod Kumar, Salman Latheef T A, Naresh Kumar A · 2025
The social networking sites of Instagram, WhatsApp, Twitter, and YouTube have led to this tremendous volume of user-generated content in text, images, and videos. Now all these elements-from marketing to direct customer engagement-become contingent upon the analysis of the emotional context of such multimodal data. This research strongly advocates an architecture that unifies NLP for text-based content and deep learning for the visual data into emotion categorization because traditional techniques for sentiment analysis cannot merge the two input modalities into meaningful and actionable insights. While EfficientNet covers the visual inputs such as images and video frames, GPT-2, optimized with LoRA, classifies the emotions within the text comments and captions. Some pipelines are included in the real-time processing modular architecture. The outputs are executed through a chatbot on an interactive frontend constructed with ReactJS. Other salient features include ensemble-based data fusion, improved pretreatment pipelines, and scalable back-end deployment. Results: This depicts a reasonably high accuracy and scale in different platforms. The overall multimodal framework would, therefore, allow organizations to extract maximum potential from the knowledge available on social media and thereby transform the process of emotion analytics.