def do_hmm(documents, split):
train, test = bifurcate(documents, split)
# train does NOT accept generators
tagger = HiddenMarkovModelTagger.train([doc.token_label_pairs() for doc in train])
results = defaultdict(list)
for doc in test:
predicted = tagger.tag(doc.tokens)
gold_labels = doc.labels
# precision =
# recall =
token_tag_pairs = nltk.pos_tag(doc.literal)
print '\n-----------\n' + gloss(token_tag_pairs)
for (token, predicted_label), gold_label in zip(predicted, gold_labels):
results[(predicted_label, gold_label)] += [token]
def do_hmm(documents, split):
train, test = bifurcate(documents, split)
# train does NOT accept generators
tagger = HiddenMarkovModelTagger.train(
[doc.token_label_pairs() for doc in train])
results = defaultdict(list)
for doc in test:
predicted = tagger.tag(doc.tokens)
gold_labels = doc.labels
# precision =
# recall =
token_tag_pairs = nltk.pos_tag(doc.literal)
print '\n-----------\n' + gloss(token_tag_pairs)
for (token,
predicted_label), gold_label in zip(predicted, gold_labels):
results[(predicted_label, gold_label)] += [token]
def __str__(self):
# spaced_tokens, spaced_labels = spaced_tokens + '\n' + spaced_labels
return gloss(self.token_pos_tag_pairs())
parser.add_argument('--train')
parser.add_argument('--tag')
parser.add_argument('--model', default='crf.model')
opts = parser.parse_args()
# This demonstrates how to obtain the version string of CRFsuite.
print 'CRFSuite v%s' % crfsuite.version()
if opts.train:
# Create a Trainer object.
trainer = Trainer()
# Read training instances from STDIN, and set them to trainer.
with open(opts.train) as lines:
for data, labels in read_svm_format(lines):
trainer.append(data, labels, 0)
trainer.save(opts.model)
else:
tagger = Tagger(opts.model)
with open(opts.tag) as lines:
for data, gold_labels in read_svm_format(lines):
predicted_labels = tagger.tag(data)
tokens = [item[0].attr for item in data]
print gloss(zip(tokens, predicted_labels, gold_labels))
# total_probability = tagger.probability(predicted_labels)
# marginals = [tagger.marginal(label, i) for i, label in enumerate(predicted_labels)]
# Output the predicted labels with their marginal probabilities.
# print '%s:%d-%f' % (y, )
parser.add_argument('--train')
parser.add_argument('--tag')
parser.add_argument('--model', default='crf.model')
opts = parser.parse_args()
# This demonstrates how to obtain the version string of CRFsuite.
print 'CRFSuite v%s' % crfsuite.version()
if opts.train:
# Create a Trainer object.
trainer = Trainer()
# Read training instances from STDIN, and set them to trainer.
with open(opts.train) as lines:
for data, labels in read_svm_format(lines):
trainer.append(data, labels, 0)
trainer.save(opts.model)
else:
tagger = Tagger(opts.model)
with open(opts.tag) as lines:
for data, gold_labels in read_svm_format(lines):
predicted_labels = tagger.tag(data)
tokens = [item[0].attr for item in data]
print gloss(zip(tokens, predicted_labels, gold_labels))
# total_probability = tagger.probability(predicted_labels)
# marginals = [tagger.marginal(label, i) for i, label in enumerate(predicted_labels)]
# Output the predicted labels with their marginal probabilities.
# print '%s:%d-%f' % (y, )
def __str__(self):
# spaced_tokens, spaced_labels = spaced_tokens + '\n' + spaced_labels
return gloss(self.token_pos_tag_pairs())