Five cancer drugs were used to test the approach, and overall, 239 users were identified with a total of 489 tweets. All of the users’ tweets related to the drugs were collapsed into a single document. The first stage was implemented using an SVM, trained and tested using manually labeled data.