As social media continues to act as a main venue for public conversation, knowing not just the general sentiment but also the polarization of viewpoints becomes increasingly important. In order to measure sentiment polarization in online conversations, this paper employs complex text extraction methods. To evaluate the spread of ideas among user groups and extract sentiment from unstructured, noisy social media data, we outline a comprehensive approach that blends machine learning and deep learning techniques. The study addresses topics including sarcasm, ambiguity, and informal language while examining crucial metrics and models for identifying sentiment polarization. According to our research, combining sentiment polarization indicators with sophisticated text extraction helps stakeholders in fields like politics, marketing, and sociological study better understand the dynamics of public opinion. This study highlights the importance of polarization, and our findings show that combining sentiment polarization indicators with better text extraction algorithms yields useful information for stakeholders in fields such as politics, marketing, and social science. This study underlines the need of identifying polarization tendencies in order to better understand public opinion dynamics and predict changes in social debates.