def test_initialize():
"""Test initialization of CRFDataset object"""
attrs = CRFDict(KEYS1)
labels = CRFDict(KEYS2)
data = CRFDataset(attrs, labels)
data_attrs = data.get_feature_list()
for i in range(len(KEYS1)):
assert KEYS1[i] == data_attrs[i]
data_labels = data.get_label_list()
for i in range(len(KEYS2)):
assert KEYS2[i] == data_labels[i]
def test_algorithms(n_samples=100, n_features=200, n_instances=10):
X = np.random.random((n_samples, n_features))
X[np.where(X < 0.8)] = 0
X = csr_matrix(X)
labels = np.random.randint(10, size=n_samples)
instances = np.linspace(0, n_samples, n_samples / n_instances + 1)[:-1]
instances = np.round(instances)
data = CRFDataset()
data.add_group_from_array(X, labels, instances)
for algorithm in ALGORITHMS:
trainer = CRFTrainer(algorithm=algorithm, quiet=True)
trainer.train(data)
def test_matrix_conversion(n_samples=50, n_features=100, n_instances=10):
"""Test conversion of csr matrix to and from CRFDataset"""
X = np.random.random((n_samples, n_features))
X[np.where(X < 0.8)] = 0
X = csr_matrix(X)
labels = np.random.randint(len(KEYS1), size=n_samples)
instances = np.linspace(0, n_samples, n_samples / n_instances + 1)[:-1]
instances = np.round(instances)
data = CRFDataset()
data.add_group_from_array(X, labels, instances)
mat = data.to_matrix()
assert_array_almost_equal(mat.toarray(), X.toarray())
from crfsuite import crfsuite_learn, CRFDataset, CRFDict, CRFTagger
n_samples = 100
n_features = 100
n_instances = 10
n_labels = 10
features_per_sample = 5
data = np.zeros((n_samples, n_features))
indices = np.random.randint(n_features, size=(n_samples, features_per_sample))
data[np.arange(n_samples)[:, None], indices] += 1
labels = np.random.randint(n_labels, size=n_samples)
instances = np.linspace(0, n_samples, n_instances + 1)
instances = np.floor(instances[:-1]).astype(int)
#crf_data = CRFDataset().add_group_from_array(data, labels, instances)
crf_data = CRFDataset().add_groups_from_files('example_files/train_small.txt')
model = crfsuite_learn(crf_data)
crf_data_test = model.get_tagging_data_from_file(
'example_files/test_small.txt')
tagger = model.get_tagger()
output = tagger.tag(crf_data_test)
