I explored sentiment extraction from tweets using fuzzy logic. Unlike data-driven models such as LSTM and CNN, my approach utilizes a rule-based system with nine rules, trapezoidal membership functions, and SOM/Centroid for the defuzzification stage. I comprehensively compared the proposed model against other methods, including Naïve Bayes and SVM using TF-IDF Vectorizer, across five datasets from Kaggle and GitHub. Moreover, I compared the proposed fuzzy system with two other articles that used CNN and the triangular fuzzy model for sentiment analysis. The reason I chose these methods was to give a thorough comparison between black-and-white box models and conclude that our approach, despite being relatively simple, is interpretable, can outperform other models most of the time, and is best when dealing with the ambiguity of human language.