When dealing with generic social media (e.g., Twitter), some researchers have first
built a classifier to identify which post was health related, before doing any further
analysis [Paul and Dredze 2011]. Or, if focusing on a specific drug, they would first filter out all the tweets that did not mention the name of the drugs [Bian et al. 2012]. For the analysis, some looked for co-occurrences of specific words to identify correlations(e.g., Scanfeld et al. [2010]).