summaryrefslogtreecommitdiff
path: root/python
diff options
context:
space:
mode:
authorVictor Chahuneau <vchahune@cs.cmu.edu>2012-11-18 15:36:16 -0500
committerVictor Chahuneau <vchahune@cs.cmu.edu>2012-11-18 15:36:16 -0500
commite9f80543bf7ff4070a1b5cdc1c3f5648addc70ce (patch)
treeb40aef0838676859f23a635317059201ee3a64e5 /python
parentc634ab440869d77769ce6d5e0f7da7bc8c36650e (diff)
Add Python tests
Diffstat (limited to 'python')
-rw-r--r--python/tests/test_decoder.py41
1 files changed, 41 insertions, 0 deletions
diff --git a/python/tests/test_decoder.py b/python/tests/test_decoder.py
new file mode 100644
index 00000000..a74e6268
--- /dev/null
+++ b/python/tests/test_decoder.py
@@ -0,0 +1,41 @@
+#coding:utf8
+import os
+import gzip
+import cdec
+import unittest
+from nose.tools import assert_almost_equals, assert_equal
+
+weights = os.path.dirname(__file__)+'/../../tests/system_tests/australia/weights'
+ref_weights = {'WordPenalty': -2.844814, 'LanguageModel': 1.0, 'PhraseModel_0': -1.066893, 'PhraseModel_1': -0.752247, 'PhraseModel_2': -0.589793, 'PassThrough': -20.0, 'Glue': 0}
+
+grammar_file = os.path.dirname(__file__)+'/../../tests/system_tests/australia/australia.scfg.gz'
+
+input_sentence = u'澳洲 是 与 北韩 有 邦交 的 少数 国家 之一 。'
+ref_output_sentence = u'australia is have diplomatic relations with north korea one of the few countries .'
+ref_f_tree = u'(S (S (S (S (X 澳洲 是)) (X (X 与 北韩) 有 邦交)) (X 的 少数 国家 之一)) (X 。))'
+ref_e_tree = u'(S (S (S (S (X australia is)) (X have diplomatic relations (X with north korea))) (X one of the few countries)) (X .))'
+ref_fvector = {'PhraseModel_2': 7.082652, 'Glue': 3.0, 'PhraseModel_0': 2.014353, 'PhraseModel_1': 8.591477}
+
+def assert_fvector_equal(vec, ref):
+ vecd = dict(vec)
+ assert_equal(set(vecd.keys()), set(ref.keys()))
+ for k, v in ref.items():
+ assert_almost_equals(vec[k], v, 6)
+
+class TestDecoder(unittest.TestCase):
+ def setUp(self):
+ self.decoder = cdec.Decoder(formalism='scfg')
+ self.decoder.read_weights(weights)
+ with gzip.open(grammar_file) as f:
+ self.grammar = f.read()
+
+ def test_weights(self):
+ assert_fvector_equal(self.decoder.weights, ref_weights)
+
+ def test_translate(self):
+ forest = self.decoder.translate(input_sentence, grammar=self.grammar)
+ assert_equal(forest.viterbi(), ref_output_sentence)
+ f_tree, e_tree = forest.viterbi_trees()
+ assert_equal(f_tree, ref_f_tree)
+ assert_equal(e_tree, ref_e_tree)
+ assert_fvector_equal(forest.viterbi_features(), ref_fvector)