def pipeline(name):
if name == 'naive_bayes_counts':
return pipelines.naive_bayes_counts()
elif name == 'naive_bayes_tfidf':
return pipelines.naive_bayes_tfidf()
elif name == 'naive_bayes_tfidf_stopwords':
return pipelines.naive_bayes_tfidf_stopwords()
elif name == 'naive_bayes_bigram':
return pipelines.naive_bayes_bigram()
elif name == 'naive_bayes_trigram':
return pipelines.naive_bayes_trigram()
elif name == 'naive_bayes_counts_lex':
return pipeline_with_lexicon.naive_bayes_counts_lex()
elif name == 'svm_libsvc_counts':
return pipelines.svm_libsvc_counts()
elif name == 'svm_libsvc_tfidf':
return pipelines.svm_libsvc_tfidf()
elif name == 'svm_libsvc_tfidf_stopwords':
return pipelines.svm_libsvc_tfidf_stopwords()
elif name == 'svm_libsvc_counts_bigram':
return pipelines.svm_libsvc_counts_bigram()
elif name == 'svm_libsvc_embed':
return pipelines.svm_libsvc_embed()
elif name == 'svm_sigmoid_embed':
return pipelines.svm_sigmoid_embed()
elif name == 'random_forest_tfidf':
return pipelines.random_forest_tfidf()
elif name == 'random_forest_tfidf_stopwords':
return pipelines.random_forest_tfidf_stopwords()
elif name == 'random_forest_embed':
return pipelines.random_forest_embed()
else:
raise ValueError(
"pipeline name is unknown. You can add a custom pipeline in 'pipelines'"
)
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_counts()
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)
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_counts()
pipe.fit(train_X, train_y)
sys_y = pipe.predict(test_X)
assert len(sys_y) == len(test_y)
def test_full_pipelines():
train_X, train_y, test_X, test_y = train_test_data()
pipes = [
pipelines.naive_bayes_counts,
pipelines.svm_libsvc_embed(),
pipelines.naive_bayes_counts_lex()
]
for pipe in pipes:
pipe = pipelines.naive_bayes_counts()
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_counts':
return pipelines.naive_bayes_counts()
elif name == 'naive_bayes_tfidf':
return pipelines.naive_bayes_tfidf()
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'"
)