def setUp(self):
"""Load data and train classifiers"""
segmentation_extractor = BaselineSegmentationFeatureExtractor()
segmentation_train_instances = (
segmentation_extractor.extract_features(inst)
for inst in load_segmentation_instances(
os.path.join(SENTENCE_SPLIT_DIR, "train.json")))
self.segmentation_dev_instances = [
segmentation_extractor.extract_features(inst)
for inst in load_segmentation_instances(
os.path.join(SENTENCE_SPLIT_DIR, "dev.json"))
]
self.segmentation_classifier = NaiveBayesClassifier(2.0)
self.segmentation_classifier.train(segmentation_train_instances)
def setUp(self) -> None:
# Create instance counter and count the instances
self.inst_counter = InstanceCounter()
feature_extractor = SegmentationTestFeatureExtractor()
self.inst_counter.count_instances(
feature_extractor.extract_features(inst)
for inst in load_segmentation_instances(
os.path.join(SENTENCE_SPLIT_DIR, "dev.json")))
def test_tuned_segmentation(self):
segmentation_extractor = TunedSegmentationFeatureExtractor()
self.assertIsNotNone(segmentation_extractor.k)
segmentation_train_instances = (
segmentation_extractor.extract_features(inst)
for inst in load_segmentation_instances(
os.path.join(SENTENCE_SPLIT_DIR, "train.json")))
self.segmentation_dev_instances = [
segmentation_extractor.extract_features(inst)
for inst in load_segmentation_instances(
os.path.join(SENTENCE_SPLIT_DIR, "dev.json"))
]
self.segmentation_classifier = NaiveBayesClassifier(
segmentation_extractor.k)
self.segmentation_classifier.train(segmentation_train_instances)
predicted, expected = self.segmentation_classifier.test(
self.segmentation_dev_instances)
acc, prec, rec, f1_score, report = classification_report(
predicted, expected, "y")
print(
f"Tuned segmentation performance for k of {segmentation_extractor.k}:"
)
print(report)
def test_sentence_split_feature_extractor(self):
"""Test feature extractor for sentence segmentation"""
label_set = frozenset(["y", "n"])
seg_feature_extractor = BaselineSegmentationFeatureExtractor()
instances = load_segmentation_instances(
os.path.join(SENTENCE_SPLIT_DIR, "dev.json"))
classification_instance = seg_feature_extractor.extract_features(
next(instances))
self.assertEqual(ClassificationInstance, type(classification_instance))
self.assertEqual(list, type(classification_instance.features))
self.assertGreaterEqual(3, len(classification_instance.features))
self.assertEqual(str, type(classification_instance.features[0]))
self.assertEqual("n", classification_instance.label)
self.assertSetEqual(
{"split_tok=.", "left_tok=D", "right_tok=Forrester"},
set(classification_instance.features),
)
for inst in instances:
classify_inst = seg_feature_extractor.extract_features(inst)
self.assertIn(classify_inst.label, label_set)