Thanks Sean you are really a font of useful information! Yes, those parameters did enable the code to run without erroring. But like you said all it did was highlight the fact that it's not a good solution and creates vast numbers of bloated objects with little to no relevant data, when all that really might be needed is a succinct text span including the medication and some form of the sig. Interestingly, as written it only captures information that's explicitly formatted as a Sig(dose form strength details). I will leave it be and write something lightweight if needed.
On Wed, Dec 4, 2024 at 7:51 PM Finan, Sean <sean.fi...@childrens.harvard.edu.invalid> wrote: > The Drug-ner module was created before my time, and I haven't ever had > reason to use it. > > I think that your oob exception is, as you guess, coming from missing or > bad parameters. You can try using the following in your piper file: > > package org.apache.ctakes.drugner.ae > set DISTANCE_ANN_TYPE=org.apache.ctakes.typesystem.type.syntax.BaseToken > set > STATUS_BOUNDARY_ANN_TYPE=org.apache.ctakes.typesystem.type.textspan.Segment > add DrugMentionAnnotator DISTANCE="10" > > Distance "10" is quoted because the code is old and doesn't accept integer > values. > > I have no idea what values would be useful, let alone optimal, for the > parameters. I couldn't find any documentation and there aren't default > values. The crash (I got it too before noticing that parameters were > required) was unexpected and not explained. > > There is a readme in the module that mentions free text and cda, but I am > not willing to do any research confirming what works or what is best. > > I hate to say it, but unless somebody reading this can offer some advice > you should probably throw up those hands and look for another solution to > extracting drug attributes. > > Definitely less interesting than sunken treasure. > > Sean > ________________________________ > From: Peter Abramowitsch <pabramowit...@gmail.com> > Sent: Wednesday, December 4, 2024 12:58 PM > To: dev@ctakes.apache.org <dev@ctakes.apache.org> > Subject: Re: Question about DrugNerAnnotator [EXTERNAL] > > * External Email - Caution * > > > Hi Sean et al > > > Problem still is though, that even when processing it creates nothing at > all due to an out of bounds exception that may be due to some parameters > that need passing to the annotator, > > but I don't see any pipeline or piper examples where these are passed. It > generates DrugMentions, then it should pass through a loop where the > attributes of the drug mention are > > transferred over to the MedicationMentions that were created upstream. > Then it deletes all the DrugMentions before exiting. Problem is that that > transfer code is never executed. > The out of bounds exception is generated here, preventing the transfer code > to execute > > int[] localSpan = getNarrativeSpansContainingGivenSpanType(jcas, > drugTokenAnt.getBegin(), iBoundaryAnnType); > > > where iBoundaryAnnType defaults to -1. It could be given the ID of a > boundary annotation through a run-time param, but I have no clue what the > basis would be for choosing a boundary type > as the whole thing's pretty large, opaque and dated. I've devoted as much > time as I can afford and am reaching the throw-up-my hands moment. I'm not > even sure if it handles free text or is > > designed only for processing CDAs > > It feels like the moment the antikythera machine was found, but less > interesting > Peter > > On Wed, Dec 4, 2024 at 4:33 PM Finan, Sean > <sean.fi...@childrens.harvard.edu.invalid> wrote: > > > Thanks for the update. We can add it a "known issues" in the wiki > > > https://urldefense.com/v3/__https://github.com/apache/ctakes/wiki__;!!NZvER7FxgEiBAiR_!vj-icS4HEyslSHEYJfEQyyi4SxnUbd59-BnMSTsAZc3YcBFmkVBBJzqPDgm6VgINUkIOJbPK7IlDetzjjEk7NJqsYGDRSbnT7A$ > . I thought that we had one at > > some point, but it may have been removed. > > > > ________________________________ > > From: Peter Abramowitsch <pabramowit...@gmail.com> > > Sent: Wednesday, December 4, 2024 10:25 AM > > To: dev@ctakes.apache.org <dev@ctakes.apache.org> > > Subject: Re: Question about DrugNerAnnotator [EXTERNAL] > > > > * External Email - Caution * > > > > > > Turns out that this problem was indeed due to the fact that package names > > had changed from v4.x.x to v5.x.x from ...drugner... to ..drug.ner.. > > But across the sources and typesystem files it had not been completely > done > > and had been left in a halfway state. By reverting the package spec of > > resources in the drugner jar > > > > from *resources/org/apache/drug/ner/**<typesystem stuff>* > > to *resources/org/apache/drugner/<**typesystem stuff>* > > > > It started working again > > > > Peter > > > > On Tue, Dec 3, 2024 at 11:50 PM Finan, Sean > > <sean.fi...@childrens.harvard.edu.invalid> wrote: > > > > > Hi Peter, > > > > > > This looks like the same problem encountered here: > > > > > > https://urldefense.com/v3/__https://github.com/apache/ctakes/issues/23__;!!NZvER7FxgEiBAiR_!uPGJp5DJ9cxqun2v3tcwAuyJAadTrq-_HtGJOt8Iw0wYntcesSesFNYRbAzIbUXbJgLuGWdlUktKeT-w9nVgFzCXhJpp9uVZHw$ > > > > > > It was fixed for -standard- ctakes workflows in 6.0.0, but if you don't > > > use the PiperFileRunner or have a custom reader then you may still see > > the > > > problem. > > > > > > The root cause is still an open issue in uima, so if anybody out there > > has > > > a general solution please share. > > > > > > https://urldefense.com/v3/__https://github.com/apache/uima-uimaj/issues/234__;!!NZvER7FxgEiBAiR_!uPGJp5DJ9cxqun2v3tcwAuyJAadTrq-_HtGJOt8Iw0wYntcesSesFNYRbAzIbUXbJgLuGWdlUktKeT-w9nVgFzCXhJoefaOEnQ$ > > > > > > > > > Sean > > > > > > ________________________________ > > > From: Peter Abramowitsch <pabramowit...@gmail.com> > > > Sent: Tuesday, December 3, 2024 3:29 PM > > > To: dev@ctakes.apache.org <dev@ctakes.apache.org> > > > Subject: Question about DrugNerAnnotator [EXTERNAL] > > > > > > * External Email - Caution * > > > > > > > > > Hi All > > > > > > If anyone has tried using the Drugner in 5.1.0 have they run into this > > > error when its process() method is run. > > > > > > message: JCas type > > > "org.apache.ctakes.drugner.type.FractionStrengthAnnotation" used in > > > Java code, but was not declared in the XML type descriptor > > > > > > The thing is that this attribute is declared in the annotator's Xml > > > Type Description, and because FractionStrengthAnnotation is also the > > > first one that the code attempts to access, I'm assuming that none of > > > the other attributes are found either. > > > attribute defs will actually be found > > > > > > <typeDescription> > > > > > > <name>org.apache.ctakes.drugner.type.FractionStrengthAnnotation</name> > > > > > > <description>The fraction portion of dosages strength determination for > > the > > > Drug NER profile.</description> > > > > > > <supertypeName>uima.tcas.Annotation</supertypeName> > > > > > > </typeDescription> > > > > > > etc > > > > > > I've looked for a rogue > > > *org.apache.ctakes.drugner.types.TypeSystem.xml* elswhere in my > > > classpath and there isn't one. But there is an issue that some bits > > > of the configuration & src of this annotator refer to the package as > > > *org.apache.ctakes.drugner*. and others as > > > *org.apache.ctakes.drug.ner.* The use of both "stringifyied package > > > names" and actual packages in the FIT architecture always confuses me. > > > This discrepancy is something that crept into certain annotators whose > > > packages changed in 5.0.x. At that time I fixed all of these so they > > > would compile, but it seems there are still issues. > > > > > > Any thoughts? > > > > > > Peter > > > > > >
