Parameters to increase precision, ways to access determinent

I am interested in getting a higher precision out of ANNIE
named entity recognition, and in using ANNIE NER in a
machine learning technique where certaintly of the answer is
important.  Does anyone know of any parameters to tell GATE
to weight precision more importantly than recall?  Is there
anyway to access a determinent, a floating point number
which tells how certain ANNIE is of annotation?

Thanks,

Debbie Duong