def __init__(self, reader, extractors, estimator, epochs, learning_rate, window_size, name_model, reader_file, optimizer):
self.reader = reader
self.extractors = extractors
self.estimator = estimator
self.label_extractor = LabelExtractor(self.reader.getPosition('LABEL'))
# estimator params
self.epochs = epochs
self.learning_rate = learning_rate
self.window_size = window_size
self.name_model = name_model
self.reader_file = reader_file
self.optimizer_type = optimizer
class Classifier(object):
def __init__(self, reader, extractors, estimator, epochs, learning_rate, window_size, name_model, reader_file, optimizer):
self.reader = reader
self.extractors = extractors
self.estimator = estimator
self.label_extractor = LabelExtractor(self.reader.getPosition('LABEL'))
# estimator params
self.epochs = epochs
self.learning_rate = learning_rate
self.window_size = window_size
self.name_model = name_model
self.reader_file = reader_file
self.optimizer_type = optimizer
def predict(self):
self.reader.load(pickle.load(open(self.reader_file)))
sentences, _ = self.reader.read()
X = []
for sentence in sentences:
feats = []
for extractor in self.extractors:
feats.append(extractor.extract(sentence, self.reader.vocabulary))
X.append([el for el in zip(*feats)])
self.estimator = self.estimator(
epochs = self.epochs,
num_labels = len(self.reader.vocabulary[self.reader.getPosition('LABEL')]),
learning_rate = self.learning_rate,
window_size = self.window_size,
num_feats= len(self.extractors),
name_model = self.name_model
)
predicted = self.estimator.predict(X)
return predicted
def train(self):
sentences, labels = self.reader.read()
X = []
y = []
for sentence, listLabels in zip(sentences, labels):
feats = []
for extractor in self.extractors:
feats.append(extractor.extract(sentence, self.reader.vocabulary))
X.append([el for el in zip(*feats)])
y.append(self.label_extractor.extract(listLabels, self.reader.vocabulary))
pickle.dump(self.reader.dump(), open(self.reader_file, 'wb'))
# splitting in train and dev
train_dataset = X[int(len(X)*0.3):]
train_labels = y[int(len(X)*0.3):]
dev_dataset = X[0:int(len(X)*0.3)]
dev_labels = y[0:int(len(X)*0.3)]
self.estimator = self.estimator(
epochs = self.epochs,
num_labels = len(self.reader.vocabulary[self.reader.getPosition('LABEL')]),
learning_rate = self.learning_rate,
window_size = self.window_size,
num_feats= len(self.extractors),
name_model = self.name_model
)
path = self.estimator.train(train_dataset, train_labels, dev_dataset, dev_labels)
return path
