def predict(self, data):
rows = []
for sentence in data:
rows.append(self.features_for_sentence(sentence))
feat_lu = lambda f: {self.vocab[item]:f[item] for item in f if item in self.vocab}
rows = [map(feat_lu, x) for x in rows]
libml.write_features(self.filename, rows, None, self.type);
libml.predict(self.filename, self.type)
labels_list = libml.read_labels(self.filename, self.type)
for t, labels in labels_list.items():
tmp = []
for sentence in data:
tmp.append([labels.pop(0) for i in range(len(sentence))])
tmp[-1] = map(lambda l: l.strip(), tmp[-1])
tmp[-1] = map(lambda l: Model.reverse_labels[int(l)], tmp[-1])
labels_list[t] = tmp
return labels_list
def train(self, data, labels):
rows = []
for sentence in data:
rows.append(self.features_for_sentence(sentence))
for row in rows:
for features in row:
for feature in features:
if feature not in self.vocab:
self.vocab[feature] = len(self.vocab) + 1
label_lu = lambda l: Model.labels[l]
labels = [map(label_lu, x) for x in labels]
feat_lu = lambda f: {self.vocab[item]:f[item] for item in f}
rows = [map(feat_lu, x) for x in rows]
libml.write_features(self.filename, rows, labels, self.type)
with open(self.filename, "w") as model:
pickle.dump(self, model)
libml.train(self.filename, self.type)