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_grid_search():
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)
params = {'clf__C': (0.1, 1)}
best_sys_y = utils.grid_search(pipelines.svm_libsvc_counts(), params,
train_X, train_y, test_X)
assert len(best_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'")