The crfppftvec module converts a simple feature vector notation to a CRF++ template file. This could be useful in case you want to dynamically configure and train a classifier using CRF++.

Simple notation:

word[0:3];pos[-2:3];bipos[-2:2];triword[-3:2];B

...where word[x,y] stands for token features from x to y, bipos[x,y] stands for bigram POS tag features between x and y, and triword[x,y] stands for trigram features between x and y. See here for more details on CRF++ feature templates.

Resulting CRF++ template:

U00:%x[0,0]
U01:%x[1,0]
U02:%x[2,0]
U03:%x[3,0]
U04:%x[-2,1]
U05:%x[-1,1]
U06:%x[0,1]
U07:%x[1,1]
U08:%x[2,1]
U09:%x[3,1]
U10:%x[-2,1]/%x[-1,1]
U11:%x[-1,1]/%x[0,1]
U12:%x[0,1]/%x[1,1]
U13:%x[1,1]/%x[2,1]
U14:%x[-3,0]/%x[-2,0]/%x[-1,0]
U15:%x[-2,0]/%x[-1,0]/%x[0,0]
U16:%x[-1,0]/%x[0,0]/%x[1,0]
U17:%x[0,0]/%x[1,0]/%x[2,0]
B

Use to_crfpp_template to convert simple notation to CRF++ template.

from crfppftvec import to_crfpp_template
ftvec = 'word[0:3];pos[-2:3];bipos[-2:2];triword[-3:2];B'
template = to_crfpp_template(ftvec)

Run installation script with:

$ python setup.py install

$ crf_learn <template_file> <train_file> <model_file>