Fig. 2.

Automated annotation using cTAKES. Left: reports were analyzed for EntityMentions, which include both disease states and anatomical terminology. Right: each EntityMention was processed using the custom NLP pipeline to identify several different language properties. These language properties include polarity (presence versus absence of the entity), subject (who does the entity pertain to?), history (is this a current or a prior condition?), and uncertainty (are we confident that this is the correct entity?)