Polarity detection is an emerging area of research in text mining. Polarity detection is observing and identifying the sentiment inclination of text, whether it is positive or negative. In this paper, a fast mode of supervised learning for polarity detection on tweets is proposed, that is using datasets available in public. The feature selection strategy ensures reduced dimensionality. The low dimension data processing on Apache Spark supports scalability for large datasets. The experiment shows that the method is supporting high scalability and efficiency. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021