This study provides a deep learning-based system for analyzing public opinion on certain topics using Twitter data. Several transformer models like BERT, DistilBERT, and RoBERTa were tested, and BERT emerged as the most successful for classification tasks. The system gets relevant tweets based on a user-specified topic, ranks them, and then does sentiment analysis on the top results. The prevalent sentiment is expressed as favorable, negative, or neutral, which has applications in a variety of industries, and is a reliable and adaptable method for measuring public opinion.