I remember there being a minimumSpan attribute in one of the xml files for the Umls lookup, it defines the minimum required span length of tokens. You could try to change this to 3 (which is the default anyway if I am not mistaken).
Regards, Tomasz ________________________________ From: Ashutosh Modi [[email protected]] Sent: Monday, November 16, 2015 3:35 PM To: user Subject: Re: difference between CVD and CPE Hi, Thanks for the reply. I figured out that there was some mistake from my side. In one of the configuration I forgot to include the relation extractor engine, so it was giving the different output. Also it was recognizing "cm" (centimeter, for e.g. in text "0.3 cm") both as disease and measurement. I explored a bit and found out that "cm" in UMLS is an abreviation for "Cutaneous Mastocytosis", which is a disease. Thanks, Ashutosh On Mon, Nov 16, 2015 at 4:12 PM, Pei Chen <[email protected]<mailto:[email protected]>> wrote: Ashutosh, That is strange. If it's the same pipeline, the results should be the same. Have you tried the CPE with only 1 doc? Could it be related to the threading issues with the LVG component? --Pei On Mon, Nov 16, 2015 at 2:43 PM, Ashutosh Modi <[email protected]<mailto:[email protected]>> wrote: > Hi, > > I am running ctakes in two different ways, one using CAS visual debugger > (CVD) and other using Collection Processing Engine (CPE). For the same text > I am getting different results from both the modes. I am using the same > engines in the both the modes. The results from CVD seem more plausible (and > correct) and output of CPE has more errors. Am I am missing something? How > can correct this? > > Please help me with this. > > Thanks, > Ashutosh
