def train(filename):
oracle = Oracle(filename)
print('oracle transitions parsing...')
oracle.search_transitions()
x = np.array(oracle.features)
y = np.array(oracle.transitions)
y = np.array([k['transition'] for k in y])
#shift:0, right:1, left:2
new_y = []
for i in range(len(y)):
if y[i] == 'shift':
new_y.append([1, 0, 0])
elif y[i] == 'right':
new_y.append([0, 1, 0])
elif y[i] == 'left':
new_y.append([0, 0, 1])
y = np.array(new_y)
print('sparse encoding of features...')
vocs, inverses = zip(*(np.unique(feature, return_inverse=True)
for feature in x.T))
x_new = np.vstack(inverses).T
x_new = np.squeeze(x_new)
clf = DecisionTreeClassifier(random_state=0)
clf.fit(x_new, y)
pickle.dump(vocs, open("vocs.p", "wb"))
pickle.dump(clf, open("model.p", "wb"))
import numpy as np
import time
from multiprocessing import Pool
from Configuration import Configuration
from Oracle import Oracle
#from Model import TBP_AS_model
filename = "../UD_French-GSD/UD_French-GSD/fr_gsd-ud-train.conllu"
#filename = "../UD_French-GSD/UD_French-GSD/test.conllu"
oracle = Oracle(filename)
oracle.search_transitions()
x = np.array(oracle.features)
y = np.array(oracle.transitions)
y = np.array([k['transition'] for k in y])
#shift:0, right:1, left:2
new_y = []
for i in range(len(y)):
if y[i] == 'shift':
new_y.append([1, 0, 0])
elif y[i] == 'right':
new_y.append([0, 1, 0])
elif y[i] == 'left':
new_y.append([0, 0, 1])
y = np.array(new_y)
x.shape, y.shape