def train(labeled_featuresets, C=1e5):
"""
:param labeled_featuresets: A list of classified featuresets,
i.e., a list of tuples ``(featureset, label)``.
"""
feat = [featureset for featureset, label in labeled_featuresets]
feature_vectorizer = MVectorizer.DictsVectorizer()
X = feature_vectorizer.fit_transform(feat)
X = Normalizer().fit_transform(X)
label_set = set( [label for featureset, label in labeled_featuresets] )
label_vectorizer = dict( [(label,num) for num,label in enumerate(label_set)] )
y = numpy.array([label_vectorizer[label] for featureset, label in labeled_featuresets])
print "Training on %d examples with %d features..."%(X.shape[0],X.shape[1]),
classifier = OneVsRestClassifier(LinearSVC(loss='l2', penalty='l2', dual=True, tol=1e-5, C=C, scale_C=True))
classifier.fit(X,y)
print "done"
return scikit_classifier(feature_vectorizer,label_vectorizer,classifier)
def predict(self, egs):
# Given a list of examples, predict their word senses
res = []
if (self.use_syntactic_features):
word_list = Syntactic_features.prepare_file(self.test_file)
syntactic = Syntactic_features.parse_stanford_output(self.test_file, word_list)
syn_index = 0
for eg in egs:
eg.word = eg.word.lower()
data,labels,pos,lesky,lesky_words = self.prepare_examples([eg], for_training=False)
# Add context words
X = self.vectorizers[eg.word].transform(data[eg.word])
# Add Parts of Speech
X_pos = self.pos_vectorizers[eg.word].transform(pos[eg.word])
X = sps.hstack((X, X_pos))
# Add Lesky Words
if self.use_lesk_words:
X_leskywords = self.lesky_words_vectorizers[eg.word].transform(lesky_words[eg.word])
X = sps.hstack((X, X_leskywords))
# Add Lesky
if self.use_lesk:
X_lesk = MVectorizer.rectangularize(lesky[eg.word])
X = sps.hstack((X, X_lesk))
# Add Syntactic dependencies
if (self.use_syntactic_features):
if all(synfeat == [] for synfeat in syntactic[syn_index]):
pass
elif (not (eg.word in self.syn_vectorizers)):
pass
else:
X_syn = self.syn_vectorizers[eg.word].transform([syntactic[syn_index]])
(x_rows,x_cols) = X.shape
(xsyn_rows,xsyn_cols) = X_syn.shape
if x_rows != xsyn_rows:
X_filler = sps.coo_matrix((x_rows-xsyn_rows,xsyn_cols))
X_syn = sps.vstack((X_syn,X_filler))
X = sps.hstack((X, X_syn))
syn_index += 1
# Add NGram model
if self.ngram_size > 0:
num_senses = self.nsenses[eg.word]
assert num_senses == len(eg.senses)
ngram_list = []
for sentence in data[eg.word]:
ngram_list.append( dict([ ( idx, self.ngram[eg.word+str(idx)].get_perplexity(sentence,True) ) for idx in range(0,num_senses) ]) )
X_ngram = MVectorizer.DictsVectorizer().fit_transform(ngram_list)
X = sps.hstack((X, X_ngram))
Y = self.classifiers[eg.word].predict(X)
senses = [0]*len(eg.senses)
for y in list(Y[0]):
senses[y] = 1
res.extend(senses)
return res
def train(self,egs):
# Trains a classifier for each word sense
data,labels,pos,lesky,lesky_words,ngram,nsenses,syntactic = self.prepare_examples(egs,verbose=True)
self.ngram = ngram
self.nsenses = nsenses
print "\nTraining on %d words"%len(data),
for word in labels.iterkeys():
sys.stdout.write(".")
sys.stdout.flush()
# Extract context features
self.vectorizers[word] = Vectorizer()
X = self.vectorizers[word].fit_transform(data[word])
# Add Parts of Speech
self.pos_vectorizers[word] = Vectorizer()
X_pos = self.pos_vectorizers[word].fit_transform(pos[word])
X = sps.hstack((X, X_pos))
# Add Lesky Words
if self.use_lesk_words:
self.lesky_words_vectorizers[word] = Vectorizer()
X_leskwords = self.lesky_words_vectorizers[word].fit_transform(lesky_words[word])
X = sps.hstack((X, X_leskwords))
# Add Lesky
if self.use_lesk:
X_lesk = MVectorizer.rectangularize(lesky[word])
X = sps.hstack((X, X_lesk))
# Add Syntactic dependencies
if (self.use_syntactic_features):
if all(synfeat == [] for synfeat in syntactic[word]):
pass
else:
self.syn_vectorizers[word] = MVectorizer.ListsVectorizer()
X_syn = self.syn_vectorizers[word].fit_transform(syntactic[word])
(x_rows,x_cols) = X.shape
(xsyn_rows,xsyn_cols) = X_syn.shape
if x_rows != xsyn_rows:
X_filler = sps.coo_matrix((x_rows-xsyn_rows,xsyn_cols))
X_syn = sps.vstack((X_syn,X_filler))
X = sps.hstack((X, X_syn))
# Add NGram model
if self.ngram_size > 0:
num_senses = self.nsenses[word]
ngram_list = []
for sentence in data[word]:
ngram_list.append( dict([ ( idx, self.ngram[word+str(idx)].get_perplexity(sentence,True) ) for idx in range(0,num_senses) ]) )
X_ngram = MVectorizer.DictsVectorizer().fit_transform(ngram_list)
X = sps.hstack((X, X_ngram))
Y = labels[word]
# Learn classifier
# self.classifiers[word] = OneVsRestClassifier(SVC(kernel='linear',scale_C=True)) #Doesn't work
self.classifiers[word] = OneVsRestClassifier(LinearSVC(loss='l2', penalty='l2', dual=False, tol=1e-3))
self.classifiers[word].fit(X,Y)
print "\nDone"
