Sentiment analysis is an interesting area of research due to the availability of sentiment data and opinion-oriented services. The efficiency and scalability of the sentiment analysis applications are important concerns as they expect accurate results in short period of time by processing a large amount of data. An efficient and scalable polarity detection method is proposed in this paper. The sequential minimal optimization with MapReduce (SMOMR) is used to achieve enhanced efficiency as well as scalability. The experiment results reveal that this method outperforms many existing methods. © 2020, Springer Nature Singapore Pte Ltd.