Sentiment analysis of text with lossless mining

Social networks are becoming more and more real with their power to influence public opinions, election outcomes, or the creation of an artificial surge in demand or supply. The continuous stream of information is valuable, but it comes with a big data problem. The question is how to mine social text at a large scale and execute machine learning algorithms to create predictive models or historical views of previous trends. This paper introduces a cyber dictionary for every user, which contains only words used in tweets - as a case study. Then, it mines all the known and unknown words by their frequency, which provides the analytic capability to run a multi-level classifier.