The exponential growth of social media platforms has resulted in a large volume of user-generated data, including opinions and sentiments expressed in various languages. This paper introduces a comprehensive framework for analyzing collective sentiment in multilingual social media feeds. It integrates web scraping, machine learning, natural language processing, and data visualization techniques. Our approach offers two primary methodologies: an English-Centric model that makes use of translation and transliteration and a Language-Specific Model approach which addresses the complexities of diverse languages. We demonstrate the framework’s value through the development of a specialized model for Romanized Nepali and the implementation of SentiSurfer, a web extension providing real-time sentiment analysis and visualization. Key features include sentiment trend analysis, geolocation-based sentiment distribution, and efficient data storage and retrieval mechanisms. By leveraging data visualization, this framework enables decision-making in fields such as business, politics, and disaster response. The proposed system offers a scale-able and efficient approach to analyzing collective sentiment in multilingual social media feeds, contributing to the development of more informed decision-making processes.