seanghay/pycrfpp

Python binding for CRF++

Python binding for CRF++

pip install pycrfpp

from pycrfpp import Tagger

# -v 3: access deep information like alpha,beta,prob
# -nN: enable nbest output. N should be >= 2
tagger = Tagger("-m ../model -v 3 -n2")

# clear internal context
tagger.clear()

# add context
tagger.add("Confidence NN")
tagger.add("in IN")
tagger.add("the DT")
tagger.add("pound NN")
tagger.add("is VBZ")
tagger.add("widely RB")
tagger.add("expected VBN")
tagger.add("to TO")
tagger.add("take VB")
tagger.add("another DT")
tagger.add("sharp JJ")
tagger.add("dive NN")
tagger.add("if IN")
tagger.add("trade NN")
tagger.add("figures NNS")
tagger.add("for IN")
tagger.add("September NNP")

print(f"column size: {tagger.xsize()}")
print(f"token size: {tagger.size()}")
print(f"tag size: {tagger.ysize()}")

print("tagset information:")

ysize = tagger.ysize()

for i in range(0, ysize - 1):
  print(f"tag {i} {tagger.yname(i)}")

tagger.parse()

print(f"conditional prob={tagger.prob()} log(Z)={tagger.Z()}")
size = tagger.size()
xsize = tagger.xsize()
for i in range(0, (size - 1)):
  for j in range(0, (xsize-1)):
    print(tagger.x(i, j) , "\t")
  
  print(tagger.y2(i) , "\t")
  print("Details")

  for j in range(0, (ysize-1)):
    print("\t" , tagger.yname(j) , "/prob=" , tagger.prob(i,j),"/alpha=" , tagger.alpha(i, j),"/beta=" , tagger.beta(i, j))
  print("\n"),

print("nbest outputs:")
for n in range(0, 9):
  if not tagger.next():
      continue
  print("nbest n=" , n , "\tconditional prob=" , tagger.prob())
  # you can access any information using tagger.y()...