def test_greedy_tagger_accuracy(self):
self.tagger = GreedyBigramTagger(0.001)
self.tagger.train(load_pos_data("test_data/train_pos.txt"))
tagged_sents = load_pos_data("test_data/100_dev.txt")
predicted, actual = self.tagger.test(tagged_sents)
accuracy = accuracy_score(actual, predicted)
print(f"GreedyBigram Accuracy: {accuracy * 100:.2f}")
self.assertLessEqual(0.953, accuracy)
class TestGreedyBigramTagger(unittest.TestCase):
def setUp(self):
self.tagger = GreedyBigramTagger(0.0)
self.tagger.train(SENTENCES_AB_XYZ)
@points(2)
def test_ab_xyz_tag1(self):
sent = ["x", "x"]
self.assertEqual(["A", "A"], self.tagger.tag_sentence(sent))
@points(2)
def test_ab_xyz_tag2(self):
sent = ["y", "y"]
self.assertEqual(["B", "B"], self.tagger.tag_sentence(sent))
@points(2)
def test_ab_xyz_tag4(self):
sent = ["x", "y"]
self.assertEqual(["A", "A"], self.tagger.tag_sentence(sent))
@points(2)
def test_ab_xyz_tag3(self):
sent = ["x", "z"]
self.assertEqual(["A", "A"], self.tagger.tag_sentence(sent))
@points(3)
def test_greedy_tagger_accuracy(self):
self.tagger = GreedyBigramTagger(0.001)
self.tagger.train(load_pos_data("test_data/train_pos.txt"))
tagged_sents = load_pos_data("test_data/100_dev.txt")
predicted, actual = self.tagger.test(tagged_sents)
accuracy = accuracy_score(actual, predicted)
print(f"GreedyBigram Accuracy: {accuracy * 100:.2f}")
self.assertLessEqual(0.953, accuracy)
class TestDeterminism(unittest.TestCase):
def setUp(self) -> None:
self.greedy_tagger = GreedyBigramTagger(0.0)
self.viterbi_tagger = ViterbiBigramTagger(0.0)
self.greedy_tagger.train(SENTENCES_NN_VB_XZY)
self.viterbi_tagger.train(SENTENCES_NN_VB_XZY)
@points(2)
def test_greedy_determinism(self):
sent = ["ghost", "pumpkin", "cider"]
self.assertEqual(["#", "#", "#"], self.greedy_tagger.tag_sentence(sent))
@points(2)
def test_viterbi_determinism(self):
sent = ["ghost", "pumpkin", "cider"]
self.assertEqual(["#", "#", "#"], self.viterbi_tagger.tag_sentence(sent))
class TestBigramSequenceProbability(unittest.TestCase):
def setUp(self):
# We test through the greedy tagger but could also do it through Viterbi
self.tagger = GreedyBigramTagger(0.0)
self.tagger.train(SENTENCES_AB_XYZ)
@points(2)
def test_prob1(self):
self.assertAlmostEqual(
-3.2188758248682006,
self.tagger.sequence_probability(["x", "y"], ["A", "A"]),
)
@points(1)
def test_prob2(self):
self.assertEqual(
float("-inf"), self.tagger.sequence_probability(["x", "y"], ["A", "B"])
)
@points(1)
def test_prob3(self):
self.assertEqual(
float("-inf"), self.tagger.sequence_probability(["x", "y"], ["B", "A"])
)
@points(2)
def test_prob4(self):
self.assertAlmostEqual(
-2.8134107167600364,
self.tagger.sequence_probability(["x", "y"], ["B", "B"]),
)
def setUp(self) -> None:
self.greedy_tagger = GreedyBigramTagger(0.0)
self.viterbi_tagger = ViterbiBigramTagger(0.0)
self.greedy_tagger.train(SENTENCES_NN_VB_XZY)
self.viterbi_tagger.train(SENTENCES_NN_VB_XZY)
def setUp(self):
self.tagger = GreedyBigramTagger(0.0)
self.tagger.train(SENTENCES_AB_XYZ)
def setUp(self):
# We test through the greedy tagger but could also do it through Viterbi
self.tagger = GreedyBigramTagger(0.0)
self.tagger.train(SENTENCES_AB_XYZ)