The clinical documents pipeline (cdpdesc/analysis_engine/AggregatePlaintextUMLSProcessor.xml) filters the drug mentions it finds, using the Orange Book. The drug NER pipeline does not do that filtering. I suspect that accounts for the differences you are seeing. The pipeline logs a message when it filters out a drug due to not finding it in the version of the Orange Book that is included within cTAKES.
If you need pointers on how to remove that filtering, I can take a look at that later today. -- James > -----Original Message----- > From: ctakes-dev-return-1156-Masanz.James=mayo....@incubator.apache.org > [mailto:ctakes-dev-return-1156-Masanz.James=mayo....@incubator.apache.org] > On Behalf Of Shady Hussein > Sent: Wednesday, February 06, 2013 4:21 AM > To: [email protected] > Subject: Re: Combining Drug and Concept annotations > > I tried it but it doesn't give me the required results. Now it annotates > the drugs as "Concept" and "MedicationEventMention" but there are still > some drugs which are not annotated while it is annotated as > "MedicationEventMention" if i used > "DrugAggregatePlaintextUMLSProcessor.xml". You can try this scenario with > this text > > DISCHARGE MEDICATIONS : > 1. Tegretol 200 mg p.o. t.i.d. > 2. Klonopin 1 mg p.o. b.i.d. > 3. Colace 100 mg p.o. b.i.d. > 4. Nicoderm 21 mg q d. topical . > 5. Zofran 24 mg p.o. q d. x 3 days . > > Here "Colace" and "Nicoderm" are not annotated but are annotated as > "MedicationEventMention" when you use > "DrugAggregatePlaintextUMLSProcessor.xml". > > Thanks for your help and looking forward to your reply > > On Wed, Feb 6, 2013 at 3:14 AM, Masanz, James J. > <[email protected]>wrote: > > > > > I'm attaching something you can try. > > > > I'm assuming you are using cTAKES 2.5. > > 1) put the attached file in cTAKESdesc\cdpdesc\analysis_engine > > 2) update > > cTAKESdesc\lookupdesc\analysis_engine\DictionaryLookupAnnotatorUMLS.xm > > l > > with your UMLS username and password > > 3) update the other copy of DictionaryLookupAnnotatorUMLS.xml in > > cTAKESdesc\drugnerdesc\analysis_engine also with your UMLS username > > and password > > > > -- James Masanz > > ________________________________________ > > From: > > ctakes-dev-return-1136-Masanz.James=mayo....@incubator.apache.org[ctak > > es-dev-return-1136-Masanz.James= [email protected]] on > > behalf of shady hussein [ [email protected]] > > Sent: Monday, February 04, 2013 1:52 PM > > To: [email protected] > > Subject: Re: Combining Drug and Concept annotations > > > > Hi Pei, > > Thanks for your reply, Yes i meant that > > DrugAggregatePlaintextUMLSProcessor return more concepts, or the > > opposite AggregatePlaintextUMLSProcessor returns the usual concepts + > > the MedicationEventMentions. I don't think it is hard to implement, as > > i think the dictionary lookup code won't change. > > > > I tried to merge the drug lookup in the lookupDB and add the drug > > annotator in the normal pipe line, but of course things are not that > > simple > > :) I don't fully understand how the dictionary look up works, > > otherwise i could do it. Maybe if you have sometime, you can guide me > > a little and i can go from there. > > > > Thanks, > > Shady > > > > On Feb 4, 2013, at 6:58 PM, "Chen, Pei" > > <[email protected]> > > wrote: > > > > > Hi Shady, > > > Just wanted to confirm: > > > Did you mean that the DrugAggregatePlaintextUMLSProcessor > > > identifies > > the same drugs, but just with more attributes (i.e. dosage, frequency, > > etc.)? > > > Or did you mean that the DrugAggregatePlaintextUMLSProcessor > > > actually > > returned more UMLSConcepts (MedicationEventMentions) the regular > > AggregatePlaintextUMLSProcessor? > > > > > > For the former, there is an outstanding Jira item to combine the 2 > > (reusing the existing lookup entries- rather than a second lookup): > > https://issues.apache.org/jira/browse/CTAKES-20 > > > > > >> -----Original Message----- > > >> From: Shady Hussein [mailto:[email protected]] > > >> Sent: Monday, February 04, 2013 5:47 AM > > >> To: [email protected] > > >> Subject: Combining Drug and Concept annotations > > >> > > >> Dear All, > > >> I discovered that cTAKES doesn't recognize all the medical > > >> entities as concepts. There is a difference between using the > > >> normal UMLS > > dictionary in > > >> "/cdpdesc/analysis_engine/AggregatePlaintextUMLSProcessor.xml" and > > >> "/drugnerdesc/analysis_engine/DrugAggregatePlaintextUMLSProcessor.x > > >> ml ". The later can detect all the drugs, while the first can't. > > >> > > >> My question now is how to combine between both of those dictionaries. > > So i > > >> can detect all the drugs and concepts mentioned in the text. I > > >> would be grateful if somebody can help me :) > > >> > > >> -- > > >> Thanks and best Regards, > > >> > > >> Shady AbdelAziz > > > > > > -- > Thanks and best Regards, > > Shady AbdelAziz
