ctakes-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Miller, Timothy" <Timothy.Mil...@childrens.harvard.edu>
Subject Re: Training for Container relation - Thyme corpus
Date Tue, 06 Sep 2016 15:25:31 GMT
Thanks for you interest. 
--xml should point to the directory where the raw anafora data sits. 
--xmi should be the directory where the xmi will be written. This is
basically a convenience -- if you rerun the eval it will check that
directory and only run the ctakes NLP pipelines if they haven't been run
--patients is the set of patient indices to use. For clinical tempeval
2016 I believe they used 1-200? 
--*-remainders is the remainders to use for train/dev/test splits, using
patient num % 8. The official split is 0,1,2,3 = train, 4,5 = dev, 6,7 =
test. The reason this is an option is because during development you
don't want to eval on test, and also because tempeval used the dev set
for testing the first year I believe.

Hope this helps.

On Tue, 2016-09-06 at 20:15 +0530, Manikandan R wrote:
> Hi all,
> I  am a newbee to Ctakes ,so please pardon my ignorance.
> I was trying to experiment with new features for container relation
> using thyme corpus and evaluate the results .
> For this I have to retrain and build model with new feature which I am
> planning to add.
> By going through the code i understood that the training and
> evaluation is done at
>  /ctakes-temporal/src/main/java/org/apache/ctakes/temporal/eval/EvaluationOfEventTimeRelations.java
> But when i try to run the above file,
> it throws error asking me to pass following option parameters
> -xml
> -xmi
> -patients
> -train-reminders
> -dev-remainders
> -test-remainders
> Any pointers regarding the explanation of these option parameters and
> how to train using Thyme corpus would be helpful

View raw message