Hi Mahesh, While there is a subject annotator and trained model, we found there were very few non-patient instances in our training data, and so the model has a difficult time finding any non-patient subjects. Unfortunately, modifying the machine learning model is not feasible, so my recommendation is, if this is important to your application, to write a rule-based annotator and add to the end of your pipeline that sets the subject attribute. Tim
On Tue, 2018-04-10 at 09:39 +0000, Mahesh Kanthaswamy wrote: > Hi Team, > > I tried running Assertion pipeline file for subject extraction. > > Its correctly annotate subject which for patient related instances > from the input narration , however it does annotate Patient’s father > as ‘Patient’ which is wrong . > > Can you please let me know what will be change I would need to > implement to correct the issue? > > > Piper: > // Add the Dependency parser for use by assertion > addDescription ClearNLPDependencyParserAE > // Add the Semantic Role Labeler parser for use by assertion > addLogged ClearNLPSemanticRoleLabelerAE > > // Add the assertion packages for class lookups > package org.apache.ctakes.assertion.medfacts > package org.apache.ctakes.assertion.attributes > add ConceptConverterAnalysisEngine > add AssertionAnalysisEngineFit > add GenericAttributeAnalysisEngine > add SubjectAttributeAnalysisEngine > > Input Narration: > > The patient had cancer and it was detected with MRI on 26th May 2017. > The patient's father had diabetes. But Patient has been prescribed > DrugXXXX for last five years. CT-SCAN of prefrontal lobe confirmed > patient had lesions in brain from last 3 months. The patient was > discharged from hospital after chemotherapy for 4 months. > > Thanks & Regards, > MAHESH KANTHASWAMY > Technical Lead | LIFE SCIENCES | Pune - SEZ > Email : [email protected] > Mobile : +91 9970406008 > > Upcoming OOO: >
