def evaluate(input_sample, gt_grams):
# char level evaluation
sample_lines = load_sample(input_sample, tokenize=False)
u_samples, b_samples, t_samples, q_samples = get_grams(sample_lines)
u_real, b_real, t_real, q_real = gt_grams
print "Unigrams: %f" % percentage_real(u_samples, u_real)
print "Bigrams: %f" % percentage_real(b_samples, b_real)
print "Trigrams: %f" % percentage_real(t_samples, t_real)
print "Quad grams: %f" % percentage_real(q_samples, q_real)
def get_gt_grams_cached(lines, dataset='training'):
grams_filename = 'true-char-ngrams.pkl'
if dataset == 'heldout':
grams_filename = 'heldout_' + grams_filename
grams_filename = FLAGS.PICKLE_PATH + '/' + grams_filename
if os.path.exists(grams_filename):
return model_and_data_serialization.load_picklized(grams_filename)
else:
grams = get_grams(lines)
model_and_data_serialization.save_picklized(grams, grams_filename)
return grams