def main():
parser = argparse.ArgumentParser(description="""Convert conllu to conll format""")
parser.add_argument('--infile', help="conllu file")
parser.add_argument('--lang', help="")
args = parser.parse_args()
#try:
header = ["proj_pred", "proj_gold", "leaf_viol_pred", "leaf_viol_gold", "posAcc", "UAS"]
if True:
vals = []
rdr = CoNLLReader()
predicted_sentences = []
gold_sentences = []
if args.infile:
gold_sentences = rdr.read_conll_u_8cols(args.infile)
numwords = sum([len(s.nodes()[1:]) for s in predicted_sentences])
#print([int(s.is_fully_projective()) for s in predicted_sentences])
for idx,s in enumerate(gold_sentences):
print(idx,s.is_fully_projective())
def main():
parser = argparse.ArgumentParser(
description="""Convert conllu to conll format""")
parser.add_argument('--predicted', help="conllu file")
parser.add_argument('--gold', help="conllu file")
parser.add_argument('--lang', help="")
args = parser.parse_args()
#try:
header = [
"proj_pred", "proj_gold", "leaf_viol_pred", "leaf_viol_gold", "posAcc",
"UAS"
]
if True:
vals = []
rdr = CoNLLReader()
predicted_sentences = []
gold_sentences = []
if args.predicted:
predicted_sentences = rdr.read_conll_u_8cols(args.predicted)
if args.gold:
gold_sentences = rdr.read_conll_u(args.gold)
numwords = sum([len(s.nodes()[1:]) for s in predicted_sentences])
#print([int(s.is_fully_projective()) for s in predicted_sentences])
proj_pred = sum(
[int(s.is_fully_projective()) for s in predicted_sentences])
proj_gold = sum([int(s.is_fully_projective()) for s in gold_sentences])
punct_non__proj_pred = sum(
[int(s.punct_proj_violations()) for s in predicted_sentences])
punct_non__proj_gold = sum(
[int(s.punct_proj_violations()) for s in gold_sentences])
leaf_violations_pred = sum(
[s.leaf_violations()[0] for s in predicted_sentences])
leaf_violations_gold = sum(
[s.leaf_violations()[0] for s in gold_sentences])
wrongPOSgoodHeadscore = wrongPOSgoodHead(predicted_sentences,
gold_sentences)
posAcc_accum = sum([
POSAcc(p, g) for p, g in zip(predicted_sentences, gold_sentences)
]) / numwords
UAS_accum = sum(
[UAS(p, g)
for p, g in zip(predicted_sentences, gold_sentences)]) / numwords
prelength = edgelengths(predicted_sentences)
goldlength = edgelengths(gold_sentences)
avgprelength = np.std(prelength)
avggoldlength = np.std(goldlength)
vals.append(wrongPOSgoodHeadscore)
vals.append(avgprelength)
vals.append(avggoldlength)
vals.append(proj_pred / len(predicted_sentences))
vals.append(proj_pred / len(predicted_sentences))
vals.append(proj_gold / len(gold_sentences))
vals.append(punct_non__proj_pred / numwords)
vals.append(punct_non__proj_gold / numwords)
vals.append(leaf_violations_pred / numwords)
vals.append(leaf_violations_gold / numwords)
vals.append(KLdivFromMACRO_POS_from_Training(predicted_sentences))
vals.append(KLdivFromMACRO_POS_from_Training(gold_sentences))
vals.append(posAcc_accum)
vals.append(UAS_accum)
lineout = " ".join([args.lang] + ["{0:.2f}".format(x) for x in vals])
#except:
# lineout = "_\t_"
print(lineout)
