Description:A optimized version of crf++0.58 with a little faster and add some postprocess for chinese segment.
Install:
a) /configure [--prefix=/xxx/xxx]
b) make
c) sudo make install
Run:
traning:
see example/chseg/train.sh
test:
see example/chseg/test.sh
You can download a pre-trained chinese model(utf8,trained by 10073281 sentences and 56397344 features) from http://pan.baidu.com/s/1eRbYelW
ChangeLog:
@2016.02.18 toshimo
1. avoid re-allocate memory frequently: it may be a little fast(about 5%+ with 50,000,000 features).
@2016.03.03 toshimo
1. add a chinese segment model(utf8)
2. add some postprocess paramters for chinese segment
@author toshimo@126.com