Hi Arron, I don't believe there is any other cTAKES annotator that will do any better. I also don't know of any current work to improve that case. From what I remember offhand, the way dosages are picked up is based on first finding the drug name, and then finding a mention of dosage nearby. So iirc there isn't a way to simply encourage more dosages to be found that you could then associate with the drug. If no one has a better answer, I'll take a look at that code on Monday to try to jog my memory more.
Wish I had better answer for you. -- James On Apr 11, 2017 9:27 AM, "Arron Lacey" <a.s.la...@swansea.ac.uk> wrote: Hi - I am running the AggregatePlainTextFastUMLSProcessor in which I am picking out drug dosage and frequency. For the most part it is very good, but I have come across a common phrase type where it trips up. It goes something like: "Lamictal 100mg in the morning and 150mg nocte" where the 100mg is picked up in str_num and str_unit, but the 150mg is not. Would any of the other analysis engines be able to improve on this, or does anyone have any suggestions how I can pick up the 150mg and attribute it to (in this case) Lamictal? I am not too bothered about the morning and night time instructions, but in the total does of (in this case) 250 mg. Is this phrase type currently being worked on for drug annotation? As I say, this seems to be quite a common phrase and wondered wether it could be addressed with the devs? Thanks - Arron.
