"but I might revisit it on a snowy afternoon this winter. " >> let's hope for lots of those and good holidays for all.
Peter On Tue, Dec 7, 2021 at 7:05 PM Finan, Sean <sean.fi...@childrens.harvard.edu> wrote: > I think that you are both correct. > > One or more preferred texts for a concept should be available if the user > created a local copy of the umls that contains one a vocabulary that > contains a preferred text. This is the first step in the process of > creating a ctakes dictionary and takes place before the ctakes dictionary > creator gui is started. I forget what the umls tool is called. > Metamorphosys? Using default settings in metamorphosys(?) should make > available a lot of preferred texts. > > I think that the ctakes dictionary creator does need to have such > vocabularies checked as src and dest. I think that the original intention > of separate src and dest options was so that you could obtain information > like synonyms and preferred text from a vocabulary without writing a column > of its codes to the dictionary. It may not be working as intended or I may > have deviated from that tactic at some point. > > A checked dest vocabulary will add a column to the dictionary database, > but you can accelerate searches in ctakes by excluding them as desired code > sources. I am probably diving too far into the weeds with that > information. > > While it is possible that the ctakes dictionary creator leaves unspecified > preferred texts unavailable for a 'good' reason, it is more likely that I > wasn't paying enough attention during implementation. It has been 2 years > since I looked at the dictionary creator (for the case-sensitive lookup), > but I might revisit it on a snowy afternoon this winter. Of course, if > anybody out there in the dev world would like to take a first crack at it > ... > > Sean > ________________________________________ > From: Peter Abramowitsch <pabramowit...@gmail.com> > Sent: Tuesday, December 7, 2021 12:34 PM > To: dev@ctakes.apache.org > Subject: Re: empty preferredText [EXTERNAL] > > * External Email - Caution * > > > I think the issue is that preferred text in the dictionary is only > populated by matches from the "dest" vocabularies and it uses *their* > preferred text. If there's no match in any of them, then it should put the > CUI's own preferred text entry in the dictionary, but it doesn't. I'm > pretty sure It's available during the dictionary creation process, but > probably not used. > > On Tue, Dec 7, 2021 at 6:22 PM Miller, Timothy < > timothy.mil...@childrens.harvard.edu> wrote: > > > OK, I thought this might be what's happening. I did check my 2021 UMLS > > release and the cui does seem to have a preferred text but I think my > > container is using an older release. For what it's worth the CUI is: > > C0360554 > > > > and a sentence that reproduces the issue in CVD with the current release > > is: > > > > 'Patient had problems tolerating oral hydrocortisone.' > > > > I will see if I can find the older UMLS release lying around. I think the > > right workaround for now is your suggestion of using the covered text. > > > > Tim > > > > > > On Tue, 2021-12-07 at 17:59 +0100, Peter Abramowitsch wrote: > > > > * External Email - Caution * > > > > > > > > Hi Tim, > > > > > > Yes, I've definitely encountered it. It happens when the concept has a > > > > CUI_TERM which has matched the text, but there is no corresponding entry > in > > > > the SNOMED or other vocab table mapping CUI to SNOMED. The obvious > choice > > > > is to use the covered text as a surrogate, but technically it could be > PHI > > > > if that matters to you. The other thing is to see if there's an MSH term > > > > that maps using the metathesaurus. If so, including MSH in your > dictionary > > > > as a src AND dest vocab will solve the problem. > > > > > > Peter > > > > > > > > On Tue, Dec 7, 2021 at 5:45 PM Miller, Timothy < > > > > <mailto:timothy.mil...@childrens.harvard.edu> > > > > timothy.mil...@childrens.harvard.edu > > > > > wrote: > > > > > > Hello, > > > > I'm using the dictionary lookup (through ctakes-web-rest) and trying to > > > > read off the preferredText that comes back as a human-readable way to > > > > display the CUI. On a very small percentage, there does not seem to be > any > > > > preferredText. Has anyone else encountered this? Is this a limitation of > > > > the underlying ontologies or a bug we can address? > > > > Tim > > > > > > >
