Figure 2
From: Mining FDA drug labels using an unsupervised learning technique - topic modeling
The distribution of the number of drugs in the 100 topics. The cutoff for topics to perform further analysis on was set at 10 drugs and is shown on the graph.