def test_trigrams(self):
training_data = ['<s>', 'sam', 'i', 'am', 'i', 'am', 'sam']
mle = MaxLikelihoodEstimator(pseudocount=1)
mle.train(zip(training_data[3:], zip(training_data,
training_data[1:])))
self.assertEqual(mle.p('i', ('<s>', 'sam')), 0.4, 3)
def test_other_distribution(self):
training_data = ['<s>', 'sam', 'i', 'am', 'i', 'am', 'sam']
mle = MaxLikelihoodEstimator()
mle.train(zip(training_data, training_data[1:]))
self.assertEqual(mle.p('sam', '<s>'), 1.0)
self.assertEqual(mle.p('i', 'sam'), 0.5)
self.assertEqual(mle.p('am', 'i'), 1.0)
self.assertEqual(mle.p('sam', 'am'), 0.5)
def test_laplace(self):
training_data = ['<s>', 'sam', 'i', 'am', 'i', 'am', 'sam']
mle = MaxLikelihoodEstimator(pseudocount=1)
mle.train(zip(training_data, training_data[1:]))
self.assertAlmostEqual(mle.p('sam', 'i'), 0.167, 3)
self.assertAlmostEqual(mle.p('i', 'i'), 0.167, 3)
self.assertAlmostEqual(mle.p('am', 'sam'), 0.167, 3)
self.assertAlmostEqual(mle.p('sam', 'i'), 0.167, 3)
self.assertAlmostEqual(mle.p('i', 'sam'), 0.333, 3)
def test_equiprobable(self):
training_data = ['<s>', 'sam', 'i', 'am']
mle = MaxLikelihoodEstimator()
mle.train(zip(training_data, training_data[1:]))
self.assertEqual(mle.p('sam', '<s>'), 1.0)
self.assertEqual(mle.p('i', 'sam'), 1.0)
self.assertEqual(mle.p('am', 'i'), 1.0)