Pei, Thank you for your response. I got all the dev tools and checked out and built ctakes from here: https://svn.apache.org/repos/asf/ctakes/tags/ctakes-3.2.1-rc1/, updated the various places with my umls credentials and ran two sentences using CVD and the AggregatePlainUMLSProcessor:
"Absence of chest pain." yields negative polarity "Chest pain absent." yields positive polarity (see the attached screen shot). Would you mind pointing me to instructions on how to "train" ctakes to pick up my example? Thank you for your help. Best, Petr On Sat, Nov 15, 2014 at 5:16 PM, Pei Chen <[email protected]> wrote: > Petr, > Which version of cTAKES are you using? < 3.2.0 or latest 3.2.1-rc1/trunk? > Both default to use a Machine Learning based polarity algorithm. If it is > missed, more training examples is probably the way to go. > The latest one uses clearTK and trained with different features and > training data so I would be curious to see if that one picks up your > examples. > > On Sat, Nov 15, 2014 at 11:04 AM, Petr Zalesky <[email protected]> > wrote: > > > I have been investigating how polarity on a sign/symptom gets set and ran > > into interesting issue. If a physician's note in a history of present > > illness (HPI) says something like: > > > > “Absence of chest pain.” > > > > “Denied chest pain.” > > > > “Chest pain resolved.” > > > > Then cTAKES picks up the term "chest pain", assigns it the correct SNOMED > > codes and sets the polarity to -1. However, some of the de-identified > > samples say: > > > > "Chest pain absent." > > > > In this case it is also picked up by cTAKES but in this case the polarity > > is set to positive one (1). I have been trying to figure out if there > is a > > way to configure cTAKES to detect that. Any suggestions? > > > -- Petr Zalesky CTO Inferscience, Inc
