ctakes-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Miller, Timothy" <Timothy.Mil...@childrens.harvard.edu>
Subject Re: SubjectClearTkAnalysisEngine not working [EXTERNAL]
Date Tue, 16 Jan 2018 19:44:20 GMT
OK, it sounds like a slight misunderstanding of what "subject" refers
to. The subject field refers to _who_ is the subject of an event.

This is important to differentiate diseases that are mentioned because
the patient is experiencing them ("pt has colon cancer") from those
that might be mentioned because a family member had them ("mother had
breast cancer").

What you're talking about sounds more like "Sections", which I think in
ctakes are called "segments". There is a regex-based section finder in
cTAKES but it is not enabled by default because it would usually need
to be customized for a given institutions notes.

Tim


On Wed, 2018-01-17 at 01:10 +0530, Ratan Sharma wrote:
> I am trying to find out something like If an entity falls in one of
> these
> category, and my understanding was subject can get me these
> information.
> 
> SUBJECT it belongs to like -
> *"Vital Signs", "BP", "Physical Examination", "Family Medical
> History",
> "Lab Results"*
> 
> Any idea how to achieve this.
> 
> 
> On Wed, Jan 17, 2018 at 1:05 AM, Miller, Timothy <
> Timothy.Miller@childrens.harvard.edu> wrote:
> 
> > 
> > What output would you like? What are you expecting?
> > 
> > This field in theory could have a few different values: patient,
> > family_member, other, donor(iirc?)
> > 
> > But in reality our training data was very skewed towards the
> > patient
> > label, and the representation we used for training is not great at
> > picking up section-wide cues that would be helpful (like a family
> > history section header). So in practice it almost always will say
> > "patient." It may occasionally get something very obvious: "Mother
> > had
> > breast cancer"
> > I don't know if it will get this exact example, it probably needs
> > to
> > look exactly like a training instance because we had very few to
> > generalize from.
> > Thanks
> > Tim
> > 
> > 
> > On Wed, 2018-01-17 at 00:57 +0530, Ratan Sharma wrote:
> > > 
> > > I am able to pull entity information for different section
> > > correctly.
> > > But
> > > facing issues when it comes to pull subject information. The
> > > subject
> > > is
> > > always pulled as "PATIENT".
> > > 
> > > I do have this added in the AssertionPipeline
> > > builder.add(
> > > SubjectCleartkAnalysisEngine.createAnnotatorDescription() );
> > > 
> > > 
> > > Here are some sample output :
> > > 
> > > Entity: 3 === Text: Blood Transfusion === Polarity: 1 ===
> > > Subject:
> > > patient
> > > === EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.SignSymptomMention
> > > Entity: 6 === Text: Blood === Polarity: 1 === Subject: patient
> > > ===
> > > EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.AnatomicalSiteMention
> > > Entity: 3 === Text: Transfusion Reaction === Polarity: 1 ===
> > > Subject:
> > > patient === EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.SignSymptomMention
> > > Entity: 5 === Text: Transfusion === Polarity: 1 === Subject:
> > > patient
> > > ===
> > > EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.ProcedureMention
> > > Entity: 2 === Text: HIV === Polarity: 1 === Subject: patient ===
> > > EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.DiseaseDisorderMention
> > > Entity: 6 === Text: Sickle Cell === Polarity: 1 === Subject:
> > > patient
> > > ===
> > > EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.AnatomicalSiteMention
> > > Entity: 2 === Text: Neurologic Disorders === Polarity: 1 ===
> > > Subject:
> > > patient === EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.DiseaseDisorderMention
> > > Entity: 2 === Text: Autoimmune Disorders === Polarity: 1 ===
> > > Subject:
> > > patient === EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.DiseaseDisorderMention
> > > Entity: 3 === Text: Autoimmune === Polarity: 1 === Subject:
> > > patient
> > > ===
> > > EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.SignSymptomMention
> > > Entity: 2 === Text: Autoimmune Disorders === Polarity: -1 ===
> > > Subject:
> > > patient === EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.DiseaseDisorderMention
> > > Entity: 3 === Text: Autoimmune === Polarity: 1 === Subject:
> > > patient
> > > ===
> > > EntityName:
> > > org.apache.ctakes.typesystem.type.textsem.SignSymptomMention
Mime
View raw message