def test_naive_bayes_pipeline():
task = of.Offenseval()
task.load(offenseval_data_dir)
train_X, train_y, test_X, test_y = utils.get_instances(
task, split_train_dev=True, proportion_train=0.1, proportion_dev=0.01)
pipe = pipelines.naive_bayes()
pipe.fit(train_X, train_y)
sys_y = pipe.predict(test_X)
assert len(sys_y) == len(test_y)
def test_hate_speech():
task = vf.VuaFormat()
task.load(hate_speech_data_dir, ['testData.csv'])
train_X, train_y, test_X, test_y = utils.get_instances(
task, split_train_dev=True, proportion_train=0.1, proportion_dev=0.01)
pipe = pipelines.naive_bayes()
pipe.fit(train_X, train_y)
sys_y = pipe.predict(test_X)
assert len(sys_y) == len(test_y)
def pipeline(name):
if name == 'naive_bayes':
return pipelines.naive_bayes()
elif name == 'svm_libsvc_counts':
return pipelines.svm_libsvc_counts()
elif name == 'svm_libsvc_tfidf':
return pipelines.svm_libsvc_tfidf()
elif name == 'svm_libsvc_embed':
return pipelines.svm_libsvc_embed()
elif name == 'svm_sigmoid_embed':
return pipelines.svm_sigmoid_embed()
else:
raise ValueError("pipeline name is unknown. You can add a custom pipeline in 'pipelines'")