ctakes-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Masanz, James J." <Masanz.Ja...@mayo.edu>
Subject cTAKES doesn't pull spatial concepts
Date Mon, 08 Jul 2013 19:02:23 GMT

I was forwarded the question below. My reply, also below, provided enough information for
the problem to be solved, so the original questioner suggested I  share the question and reply


 Apache cTAKES is configured to annotate named entities only within certain semantic types
(that is, only concepts with certain UMLS TUI values)
So if "right occipital region"  is a UMLS concept  not within a semantic type cTAKES annotates,
at minimum, you need to add the semantic type(s) you want included to be within one of the
LookupDesc*xml files  (which one depends on which pipeline you are running)
For example, LookupDesc_Db.xml contains this line:
<property key="anatomicalSiteTuis" value="T021,T022,T023,T024,T025,T026,T029,T030"/>
which shows that only those TUIs are considered when annotating anatomical sites

So it might be a case of adding a TUI to that list.

On the other hand it could just be either of these:
- maybe SNOMED-CT doesn't distinguish between left and right occipital regions?
- or maybe it does, but UMLS concepts do not - it is possible in such a case that 2 or more
SNOMED-CT terms map to a single UMLS concept if the SNOMED-CT terms are more specific.

Which leads to the question - specifically, which UMLS concept(s) were you expecting cTAKES
to find/return?



I am using the CAS Visual Debugger to pull UMLS concepts out of my text using the AggregatePlaintextUMLSProcessor
analysis engine, and it works great.  However, I noticed that it doesn't pull spacial concepts
-- for instance, in the phrase "right occipital region" it only identifies "occipital region"
as a concept.  It would be nice to have the location and unique identifier for spacial concepts.
 Is there a place in the CVD gui to change the filters so that it picks up things like spacial
concepts?  Do I have to go with another analysis engine?


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message