Improving cTAKES performance "out of the box"

[David Kincaid <ki...@gmail.com>]

I've been starting to work with cTAKES recently and trying out the clinical
pipeline on some documents. I was wondering what were some of the first
things that people are modifying to improve the performance? When I look at
the annotations after running some discharge documents through, the
performance on negation, history of, concept detection (medications,
signs/symptoms, disease/disorder, etc), etc are not very good. Are there
newer components or models that people are plugging in?

I've only just started working with it (although I have been following the
project for some time) as a side project/proof of concept to try to get
some funding for a bigger project. This looks like a very good starting
point and I'm hoping to get a chance to work with it a lot more. I
appreciate any advice the group might have.

Thanks,

Dave Kincaid