As I mentioned in the article but admittedly didn't go into depth on...
Apache ctakes will be used to analyze and extract as much data as possible from the free form journals.
This will then be used to create a searchable graph across as many dimensions of data as we can extract.
From there we will use statistical analysis to find correlates.
You can do interesting things with graph databases that you can't do with rdbms. Such as ask really weird freeform questions like how many people ate Doritos on Monday and reported feelings of depression on Wednesday then took Tylenol vs Ibuprofen vs Water and felt better by Thursday?
These types of things are only possible with graph based analysis and freeform extraction. And who knows? We might uncover something completely unexpected like Type 1 diabetics respond better to water than Naproxen Sodium. And yes that's an actual theory I would like to test because Type 1 diabetes is an Immune system malfunction where the immune system is already misidentifying things.