def generate_chomsky(times=5, line_length=72):
parts = []
for part in (leadins, subjects, verbs, objects):
phraselist = list(map(str.strip, part.splitlines()))
random.shuffle(phraselist)
parts.append(phraselist)
output = chain(*islice(izip(*parts), 0, times))
print(textwrap.fill(" ".join(output), line_length))
def accuracy(self, feature_sets):
X, y = list(compat.izip(*feature_sets))
classifications = self.classify_many(X)
count = 0
print 'Y LENGTH: ' + str(len(y))
print 'CLASSIFICATIONS LENGTH: ' + str(len(classifications))
for i in range(0, len(classifications)):
if classifications[i] == y[i]:
count += 1
accuracy = (count * 1.0) / len(y)
return accuracy
def train(self, feature_sets):
X, y = list(compat.izip(*feature_sets))
X = self.vectorizer.fit_transform(X)
y = self.encoder.fit_transform(y)
if self.thetas == None:
self.thetas = sparse.rand(X.shape[1], 1, 1, 'csr')
cost = self.compute_cost(X, y)
print cost
self.gradient_descent(X, y)
def test( self, test_sequence, verbose = False, **kwargs ):
"""
Tests the HiddenMarkovModelTagger instance.
:param test_sequence: a sequence of labeled test instances
:type test_sequence: list(list)
:param verbose: boolean flag indicating whether training should be
verbose or include printed output
:type verbose: bool
"""
def words( sent ):
return [ word for (word, tag) in sent ]
def tags( sent ):
return [ tag for (word, tag) in sent ]
def flatten( seq ):
return list( itertools.chain( *seq ) )
test_sequence = self._transform( test_sequence )
predicted_sequence = list( imap( self._tag, imap( words, test_sequence ) ) )
if verbose:
for test_sent, predicted_sent in izip( test_sequence, predicted_sequence ):
print( 'Test:',
' '.join( '%s/%s' % (token, tag)
for (token, tag) in test_sent ) )
print( )
print( 'Untagged:',
' '.join( "%s" % token for (token, tag) in test_sent ) )
print( )
print( 'HMM-tagged:',
' '.join( '%s/%s' % (token, tag)
for (token, tag) in predicted_sent ) )
print( )
print( 'Entropy:',
self.entropy( [ (token, None) for
(token, tag) in predicted_sent ] ) )
print( )
print( '-' * 60 )
test_tags = flatten( imap( tags, test_sequence ) )
predicted_tags = flatten( imap( tags, predicted_sequence ) )
acc = accuracy( test_tags, predicted_tags )
count = sum( len( sent ) for sent in test_sequence )
print( 'accuracy over %d tokens: %.2f' % (count, acc * 100) )
def train(self, labeled_featuresets):
"""
Train (fit) the scikit-learn estimator.
:param labeled_featuresets: A list of ``(featureset, label)``
where each ``featureset`` is a dict mapping strings to either
numbers, booleans or strings.
"""
X, y = list(compat.izip(*labeled_featuresets))
X = self._vectorizer.fit_transform(X)
y = self._encoder.fit_transform(y)
self._clf.fit(X, y)
return self
def accuracy(reference, test):
"""
Given a list of reference values and a corresponding list of test
values, return the fraction of corresponding values that are
equal. In particular, return the fraction of indices
``0<i<=len(test)`` such that ``test[i] == reference[i]``.
:type reference: list
:param reference: An ordered list of reference values.
:type test: list
:param test: A list of values to compare against the corresponding
reference values.
:raise ValueError: If ``reference`` and ``length`` do not have the
same length.
"""
if len(reference) != len(test):
raise ValueError("Lists must have the same length.")
return sum(x == y for x, y in izip(reference, test)) / len(test)
def are_files_identical(filename1, filename2, debug=False):
"""
Compare two files, ignoring carriage returns.
"""
with open(filename1, "rb") as fileA:
with open(filename2, "rb") as fileB:
result = True
for lineA, lineB in izip(sorted(fileA.readlines()),
sorted(fileB.readlines())):
if lineA.strip() != lineB.strip():
if debug:
print("Error while comparing files. " +
"First difference at line below.")
print("=> Output file line: {0}".format(lineA))
print("=> Refer. file line: {0}".format(lineB))
result = False
break
return result
def accuracy(reference, test):
"""
Given a list of reference values and a corresponding list of test
values, return the fraction of corresponding values that are
equal. In particular, return the fraction of indices
``0<i<=len(test)`` such that ``test[i] == reference[i]``.
:type reference: list
:param reference: An ordered list of reference values.
:type test: list
:param test: A list of values to compare against the corresponding
reference values.
:raise ValueError: If ``reference`` and ``length`` do not have the
same length.
"""
if len(reference) != len(test):
raise ValueError("Lists must have the same length.")
return float(sum(x == y for x, y in izip(reference, test))) / len(test)
def are_files_identical(filename1, filename2, debug=False):
"""
Compare two files, ignoring carriage returns.
"""
with open(filename1, "rb") as fileA:
with open(filename2, "rb") as fileB:
result = True
for lineA, lineB in izip(sorted(fileA.readlines()),
sorted(fileB.readlines())):
if lineA.strip() != lineB.strip():
if debug:
print("Error while comparing files. " +
"First difference at line below.")
print("=> Output file line: {0}".format(lineA))
print("=> Refer. file line: {0}".format(lineB))
result = False
break
return result
def log_likelihood(reference, test):
"""
Given a list of reference values and a corresponding list of test
probability distributions, return the average log likelihood of
the reference values, given the probability distributions.
:param reference: A list of reference values
:type reference: list
:param test: A list of probability distributions over values to
compare against the corresponding reference values.
:type test: list(ProbDistI)
"""
if len(reference) != len(test):
raise ValueError("Lists must have the same length.")
# Return the average value of dist.logprob(val).
total_likelihood = sum(dist.logprob(val) for (val, dist) in izip(reference, test))
return total_likelihood / len(reference)
def log_likelihood(reference, test):
"""
Given a list of reference values and a corresponding list of test
probability distributions, return the average log likelihood of
the reference values, given the probability distributions.
:param reference: A list of reference values
:type reference: list
:param test: A list of probability distributions over values to
compare against the corresponding reference values.
:type test: list(ProbDistI)
"""
if len(reference) != len(test):
raise ValueError("Lists must have the same length.")
# Return the average value of dist.logprob(val).
total_likelihood = sum(
dist.logprob(val) for (val, dist) in izip(reference, test))
return total_likelihood / len(reference)
def train(self, labeled_featuresets):
X, y = list(izip(*labeled_featuresets))
X = self._vectorizer.fit_transform(X)
y = self._encoder.fit_transform(y)
row, column = X.shape
print 'row:'+str(row)
print 'column:'+str(column)
print y.shape
print X
print y
with open('matrix', 'w') as f:
for i in xrange(0, row):
f.write(str(y[i]) + ' ')
for j in xrange(0, column):
f.write(str(int(X[i,j])) + ' ')
f.write('\n')
f.close()
self._clf.fit(X, y)
return self
def _tag(self, unlabeled_sequence):
path = self._best_path(unlabeled_sequence)
return list(izip(unlabeled_sequence, path))
def _tag(self, unlabeled_sequence):
path = self._best_path(unlabeled_sequence)
return list(izip(unlabeled_sequence, path))
def matrix_dimension(feature_set):
vectorizer = DictVectorizer(dtype=float, sparse=True)
X, _ = list(compat.izip(*feature_set))
X = vectorizer.fit_transform(X)
return X.toarray().shape
affix_tagger = EntropyAffixTagger(train)
unigram_tagger = EntropyUnigramTagger(train)
taggers = [unigram_tagger, affix_tagger]
tagger = EntropyVotingTagger(taggers, max_entropy=80)
from nltk.tag import untag
untagged_test = [untag(x) for x in dev]
tagged_sents_uni_affix = unigram_affix_backoff.tag_sents(untagged_test)
tagged_sents_entr = tagger.tag_sents(untagged_test)
affix_mistake = 0
unigram_mistake = 0
overall_mistakes = 0
print "len of dev: ", len(dev)
for tagged_reference_sent, tagged_uni_affix_sent, tagged_entropy_sent in izip(dev, tagged_sents_uni_affix,
tagged_sents_entr):
# import pdb;pdb.set_trace()
for tagged_reference, tagged_uni_affix, tagged_entropy in izip(tagged_reference_sent, tagged_uni_affix_sent,
tagged_entropy_sent):
if tagged_uni_affix[1] != tagged_entropy[1]:
overall_mistakes += 1
print "WE GOT MATCH!"
print "Word = ", tagged_reference[0]
print "real tag ", tagged_reference[1]
print "backoff tag ", tagged_uni_affix[1]
print "entropy tag ", tagged_entropy[1]
for t in tagger._taggers:
# import pdb
# pdb.set_trace()
print "Entropy for tagger ", t.__class__.__name__, " ", t.entropy(tagged_reference[0])
