def train(self, filename):
data = self.load_data(filename)
m = MaxentModel()
m.begin_add_event()
for word, history in data:
features = self.generate_features(word, history)
m.add_event(features, word)
m.end_add_event()
m.train()
self.model = m
def main():
m = MaxentModel()
data = load_data()
m.begin_add_event()
for vals, label in data:
features = extract_features(vals)
m.add_event(features, label)
m.end_add_event()
m.train()
print 'Trained'
test_instance = [6.7, 3.1, 5.6, 2.4]
test_features = extract_features(test_instance)
result = m.eval_all(test_features)
print 'Result'
print result
'SHS.RELIG': 'Religions',
'SHS.SCIPO': 'Political science',
'SHS.SOCIO': 'Sociology',
'SHS.STAT': 'Methods and statistics',
'SDV': 'Life sciences',
'SPI': 'Engineering sciences',
'STAT': 'Statistics',
'QFIN': 'Economy and quantitative finance',
'OTHER': 'Other',
}
possible_labels = readable_descriptions.keys()
print "Loading model..."
model = MaxentModel()
model.load(str('data/model.maxent'))
print "Done."
print "Loading dict"
with open('data/dict-hal.pkl', 'rb') as f:
dct = cPickle.load(f)
corpus = HALCorpus(dct=dct)
print "Done."
def lookup_category(probas):
readable_list = []
for lst in probas:
if lst[0] in readable_descriptions:
readable_list.append(
{'code':lst[0],
'proba':lst[1],
def load(self, filename):
"""Loads a maxent model from disk."""
self.model = MaxentModel()
self.model.load(filename)
class Maxent(Model):
def __init__(self, history_length=30):
Model.__init__(self)
self.history_length = history_length
def get_probability(self, word, history):
features = self.generate_features(word, history)
prob_dist = self.model.eval_all(features)
for tag, prob in prob_dist:
if tag == word:
return prob
return 0.0
def load_data(self, filename):
words = [line.strip() for line in open(filename, 'r')]
for index in xrange(len(words)):
word = words[index]
start_index = index - self.history_length \
if index - self.history_length > 0 else 0 # look at the most recent N tags
history = words[start_index:index]
yield word, history
def train(self, filename):
data = self.load_data(filename)
m = MaxentModel()
m.begin_add_event()
for word, history in data:
features = self.generate_features(word, history)
m.add_event(features, word)
m.end_add_event()
m.train()
self.model = m
def generate_features(self, word, history):
history = self.pad_history(history)
features = Feature.eval(word, history)
return features
def pad_history(self, history):
"""Pads 'START' tag if history length is less than history_length"""
if len(history) < self.history_length:
history = ['START'] * (self.history_length - len(history)) + history
return history
def save(self, filename):
"""Saves the model to a file. Just calls the internal model's
save function. We discard the value of history_length and
feature_funcs. We assume history_length will always be the same,
and we just reload the feature_funcs.
"""
self.model.save(filename)
def load(self, filename):
"""Loads a maxent model from disk."""
self.model = MaxentModel()
self.model.load(filename)