PageRenderTime 56ms CodeModel.GetById 27ms RepoModel.GetById 0ms app.codeStats 0ms

/test/test_nl.py

https://gitlab.com/github-cloud-corporation/pattern
Python | 273 lines | 270 code | 2 blank | 1 comment | 1 complexity | 19a5146666a867094d7b8d7fd4e1685b MD5 | raw file
    

  1. # -*- coding: utf-8 -*-
    2. 

  2. import os, sys; sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
    3. 

  3. import unittest
    4. 

  4. import subprocess
    5. 

  5. 

  6. from pattern import nl
    7. 

  7. 

  8. try:
    9. 

  9.     PATH = os.path.dirname(os.path.realpath(__file__))
    10. 

  10. except:
    11. 

  11.     PATH = ""
    12. 

  12. 

  13. #---------------------------------------------------------------------------------------------------
    14. 

  14. 

  15. class TestInflection(unittest.TestCase):
    16. 

  16. 

  17.     def setUp(self):
    18. 

  18.         pass
    19. 

  19.         
    20. 

  20.     def test_pluralize(self):
    21. 

  21.         # Assert "auto's" as plural of "auto".
    22. 

  22.         self.assertEqual("auto's", nl.inflect.pluralize("auto"))
    23. 

  23.         # Assert the accuracy of the pluralization algorithm.
    24. 

  24.         from pattern.db import Datasheet
    25. 

  25.         i, n = 0, 0
    26. 

  26.         for pred, attr, sg, pl in Datasheet.load(os.path.join(PATH, "corpora", "wordforms-nl-celex.csv")):
    27. 

  27.             if nl.pluralize(sg) == pl:
    28. 

  28.                 i +=1
    29. 

  29.             n += 1
    30. 

  30.         self.assertTrue(float(i) / n > 0.74)
    31. 

  31.         print("pattern.nl.pluralize()")
    32. 

  32.         
    33. 

  33.     def test_singularize(self):
    34. 

  34.         # Assert the accuracy of the singularization algorithm.
    35. 

  35.         from pattern.db import Datasheet
    36. 

  36.         i, n = 0, 0
    37. 

  37.         for pred, attr, sg, pl in Datasheet.load(os.path.join(PATH, "corpora", "wordforms-nl-celex.csv")):
    38. 

  38.             if nl.singularize(pl) == sg:
    39. 

  39.                 i +=1
    40. 

  40.             n += 1
    41. 

  41.         self.assertTrue(float(i) / n > 0.88)
    42. 

  42.         print("pattern.nl.singularize()")
    43. 

  43. 

  44.     def test_attributive(self):
    45. 

  45.         # Assert the accuracy of the attributive algorithm ("fel" => "felle").
    46. 

  46.         from pattern.db import Datasheet
    47. 

  47.         i, n = 0, 0
    48. 

  48.         for pred, attr, sg, pl in Datasheet.load(os.path.join(PATH, "corpora", "wordforms-nl-celex.csv")):
    49. 

  49.             if nl.attributive(pred) == attr:
    50. 

  50.                 i +=1
    51. 

  51.             n += 1
    52. 

  52.         self.assertTrue(float(i) / n > 0.96)
    53. 

  53.         print("pattern.nl.attributive()")
    54. 

  54.         
    55. 

  55.     def test_predicative(self):
    56. 

  56.         # Assert the accuracy of the predicative algorithm ("felle" => "fel").
    57. 

  57.         from pattern.db import Datasheet
    58. 

  58.         i, n = 0, 0
    59. 

  59.         for pred, attr, sg, pl in Datasheet.load(os.path.join(PATH, "corpora", "wordforms-nl-celex.csv")):
    60. 

  60.             if nl.predicative(attr) == pred:
    61. 

  61.                 i +=1
    62. 

  62.             n += 1
    63. 

  63.         self.assertTrue(float(i) / n > 0.96)
    64. 

  64.         print("pattern.nl.predicative()")
    65. 

  65. 

  66.     def test_find_lemma(self):
    67. 

  67.         # Assert the accuracy of the verb lemmatization algorithm.
    68. 

  68.         # Note: the accuracy is higher (90%) when measured on CELEX word forms
    69. 

  69.         # (presumably because nl.inflect.verbs has high percentage irregular verbs).
    70. 

  70.         i, n = 0, 0
    71. 

  71.         for v1, v2 in nl.inflect.verbs.inflections.items():
    72. 

  72.             if nl.inflect.verbs.find_lemma(v1) == v2: 
    73. 

  73.                 i += 1
    74. 

  74.             n += 1
    75. 

  75.         self.assertTrue(float(i) / n > 0.83)
    76. 

  76.         print("pattern.nl.inflect.verbs.find_lemma()")
    77. 

  77.         
    78. 

  78.     def test_find_lexeme(self):
    79. 

  79.         # Assert the accuracy of the verb conjugation algorithm.
    80. 

  80.         i, n = 0, 0
    81. 

  81.         for v, lexeme1 in nl.inflect.verbs.infinitives.items():
    82. 

  82.             lexeme2 = nl.inflect.verbs.find_lexeme(v)
    83. 

  83.             for j in range(len(lexeme2)):
    84. 

  84.                 if lexeme1[j] == lexeme2[j] or \
    85. 

  85.                    lexeme1[j] == "" and \
    86. 

  86.                    lexeme1[j > 5 and 10 or 0] == lexeme2[j]:
    87. 

  87.                     i += 1
    88. 

  88.                 n += 1
    89. 

  89.         self.assertTrue(float(i) / n > 0.79)
    90. 

  90.         print("pattern.nl.inflect.verbs.find_lexeme()")
    91. 

  91. 

  92.     def test_conjugate(self):
    93. 

  93.         # Assert different tenses with different conjugations.
    94. 

  94.         for (v1, v2, tense) in (
    95. 

  95.           ("zijn",  "zijn",     nl.INFINITIVE),
    96. 

  96.           ("zijn",  "ben",     (nl.PRESENT, 1, nl.SINGULAR)),
    97. 

  97.           ("zijn",  "bent",    (nl.PRESENT, 2, nl.SINGULAR)),
    98. 

  98.           ("zijn",  "is",      (nl.PRESENT, 3, nl.SINGULAR)),
    99. 

  99.           ("zijn",  "zijn",    (nl.PRESENT, 0, nl.PLURAL)),
    100. 

  100.           ("zijn",  "zijnd",   (nl.PRESENT + nl.PARTICIPLE,)),
    101. 

  101.           ("zijn",  "was",     (nl.PAST, 1, nl.SINGULAR)),
    102. 

  102.           ("zijn",  "was",     (nl.PAST, 2, nl.SINGULAR)),
    103. 

  103.           ("zijn",  "was",     (nl.PAST, 3, nl.SINGULAR)),
    104. 

  104.           ("zijn",  "waren",   (nl.PAST, 0, nl.PLURAL)),
    105. 

  105.           ("zijn",  "was",     (nl.PAST, 0, None)),
    106. 

  106.           ("zijn",  "geweest", (nl.PAST + nl.PARTICIPLE,)),
    107. 

  107.           ("had",   "hebben",   "inf"),
    108. 

  108.           ("had",   "heb",      "1sg"),
    109. 

  109.           ("had",   "hebt",     "2sg"),
    110. 

  110.           ("had",   "heeft",    "3sg"),
    111. 

  111.           ("had",   "hebben",   "pl"),
    112. 

  112.           ("had",   "hebbend",  "part"),
    113. 

  113.           ("heeft", "had",      "1sgp"),
    114. 

  114.           ("heeft", "had",      "2sgp"),
    115. 

  115.           ("heeft", "had",      "3sgp"),
    116. 

  116.           ("heeft", "hadden",   "ppl"),
    117. 

  117.           ("heeft", "had",      "p"),
    118. 

  118.           ("heeft", "gehad",    "ppart"),
    119. 

  119.           ("smsen", "smste",    "3sgp")):
    120. 

  120.             self.assertEqual(nl.conjugate(v1, tense), v2)
    121. 

  121.         print("pattern.nl.conjugate()")
    122. 

  122. 

  123.     def test_lexeme(self):
    124. 

  124.         # Assert all inflections of "zijn".
    125. 

  125.         v = nl.lexeme("zijn")
    126. 

  126.         self.assertEqual(v, [
    127. 

  127.             "zijn", "ben", "bent", "is", "zijnd", "waren", "was", "geweest"
    128. 

  128.         ])
    129. 

  129.         print("pattern.nl.inflect.lexeme()")
    130. 

  130. 

  131.     def test_tenses(self):
    132. 

  132.         # Assert tense recognition.
    133. 

  133.         self.assertTrue((nl.PRESENT, 3, "sg") in nl.tenses("is"))
    134. 

  134.         self.assertTrue("3sg" in nl.tenses("is"))
    135. 

  135.         print("pattern.nl.tenses()")
    136. 

  136. 

  137. #---------------------------------------------------------------------------------------------------
    138. 

  138. 

  139. class TestParser(unittest.TestCase):
    140. 

  140.     
    141. 

  141.     def setUp(self):
    142. 

  142.         pass
    143. 

  143.    
    144. 

  144.     def test_wotan2penntreebank(self):
    145. 

  145.         # Assert tag translation.
    146. 

  146.         for penntreebank, wotan in (
    147. 

  147.           ("NNP",  "N(eigen,ev,neut)"),
    148. 

  148.           ("NNPS", "N(eigen,mv,neut)"),
    149. 

  149.           ("NN",   "N(soort,ev,neut)"),
    150. 

  150.           ("NNS",  "N(soort,mv,neut)"),
    151. 

  151.           ("VBZ",  "V(refl,ott,3,ev)"),
    152. 

  152.           ("VBP",  "V(intrans,ott,1_of_2_of_3,mv)"),
    153. 

  153.           ("VBD",  "V(trans,ovt,1_of_2_of_3,mv)"),
    154. 

  154.           ("VBN",  "V(trans,verl_dw,onverv)"),
    155. 

  155.           ("VBG",  "V(intrans,teg_dw,onverv)"),
    156. 

  156.           ("VB",   "V(intrans,inf)"),
    157. 

  157.           ("MD",   "V(hulp_of_kopp,ott,3,ev)"),
    158. 

  158.           ("JJ",   "Adj(attr,stell,onverv)"),
    159. 

  159.           ("JJR",  "Adj(adv,vergr,onverv)"),
    160. 

  160.           ("JJS",  "Adj(attr,overtr,verv_neut)"),
    161. 

  161.           ("RP",   "Adv(deel_v)"),
    162. 

  162.           ("RB",   "Adv(gew,geen_func,stell,onverv)"),
    163. 

  163.           ("DT",   "Art(bep,zijd_of_mv,neut)"),
    164. 

  164.           ("CC",   "Conj(neven)"),
    165. 

  165.           ("CD",   "Num(hoofd,bep,zelfst,onverv)"),
    166. 

  166.           ("TO",   "Prep(voor_inf)"),
    167. 

  167.           ("IN",   "Prep(voor)"),
    168. 

  168.           ("PRP",  "Pron(onbep,neut,attr)"),
    169. 

  169.           ("PRP$", "Pron(bez,2,ev,neut,attr)"),
    170. 

  170.           (",",    "Punc(komma)"),
    171. 

  171.           ("(",    "Punc(haak_open)"),
    172. 

  172.           (")",    "Punc(haak_sluit)"),
    173. 

  173.           (".",    "Punc(punt)"),
    174. 

  174.           ("UH",   "Int"),
    175. 

  175.           ("SYM",  "Misc(symbool)")):
    176. 

  176.             self.assertEqual(nl.wotan2penntreebank("", wotan)[1], penntreebank)
    177. 

  177.         print("pattern.nl.wotan2penntreebank()")
    178. 

  178.         
    179. 

  179.     def test_find_lemmata(self):
    180. 

  180.         # Assert lemmata for nouns and verbs.
    181. 

  181.         v = nl.parser.find_lemmata([["katten", "NNS"], ["droegen", "VBD"], ["hoeden", "NNS"]])
    182. 

  182.         self.assertEqual(v, [
    183. 

  183.             ["katten", "NNS", "kat"], 
    184. 

  184.             ["droegen", "VBD", "dragen"], 
    185. 

  185.             ["hoeden", "NNS", "hoed"]])
    186. 

  186.         print("pattern.nl.parser.find_lemmata()")
    187. 

  187.     
    188. 

  188.     def test_parse(self):
    189. 

  189.         # Assert parsed output with Penn Treebank II tags (slash-formatted).
    190. 

  190.         # 1) "de zwarte kat" is a noun phrase, "op de mat" is a prepositional noun phrase.
    191. 

  191.         v = nl.parser.parse("De zwarte kat zat op de mat.")
    192. 

  192.         self.assertEqual(v,
    193. 

  193.             "De/DT/B-NP/O zwarte/JJ/I-NP/O kat/NN/I-NP/O " + \
    194. 

  194.             "zat/VBD/B-VP/O " + \
    195. 

  195.             "op/IN/B-PP/B-PNP de/DT/B-NP/I-PNP mat/NN/I-NP/I-PNP ././O/O"
    196. 

  196.         )
    197. 

  197.         # 2) "jaagt" and "vogels" lemmata are "jagen" and "vogel".
    198. 

  198.         v = nl.parser.parse("De zwarte kat jaagt op vogels.", lemmata=True)
    199. 

  199.         self.assertEqual(v,
    200. 

  200.             "De/DT/B-NP/O/de zwarte/JJ/I-NP/O/zwart kat/NN/I-NP/O/kat " + \
    201. 

  201.             "jaagt/VBZ/B-VP/O/jagen " + \
    202. 

  202.             "op/IN/B-PP/B-PNP/op vogels/NNS/B-NP/I-PNP/vogel ././O/O/."
    203. 

  203.         )
    204. 

  204.         # Assert the accuracy of the Dutch tagger.
    205. 

  205.         i, n = 0, 0
    206. 

  206.         for sentence in open(os.path.join(PATH, "corpora", "tagged-nl-twnc.txt")).readlines():
    207. 

  207.             sentence = sentence.decode("utf-8").strip()
    208. 

  208.             s1 = [w.split("/") for w in sentence.split(" ")]
    209. 

  209.             s1 = [nl.wotan2penntreebank(w, tag) for w, tag in s1]
    210. 

  210.             s2 = [[w for w, pos in s1]]
    211. 

  211.             s2 = nl.parse(s2, tokenize=False)
    212. 

  212.             s2 = [w.split("/") for w in s2.split(" ")]
    213. 

  213.             for j in range(len(s1)):
    214. 

  214.                 if s1[j][1] == s2[j][1]:
    215. 

  215.                     i += 1
    216. 

  216.                 n += 1
    217. 

  217.         self.assertTrue(float(i) / n > 0.90)
    218. 

  218.         print("pattern.nl.parser.parse()")
    219. 

  219. 

  220.     def test_tag(self):
    221. 

  221.         # Assert [("zwarte", "JJ"), ("panters", "NNS")].
    222. 

  222.         v = nl.tag("zwarte panters")
    223. 

  223.         self.assertEqual(v, [("zwarte", "JJ"), ("panters", "NNS")])
    224. 

  224.         print("pattern.nl.tag()")
    225. 

  225.     
    226. 

  226.     def test_command_line(self):
    227. 

  227.         # Assert parsed output from the command-line (example from the documentation).
    228. 

  228.         p = ["python", "-m", "pattern.nl", "-s", "Leuke kat.", "-OTCRL"]
    229. 

  229.         p = subprocess.Popen(p, stdout=subprocess.PIPE)
    230. 

  230.         p.wait()
    231. 

  231.         v = p.stdout.read()
    232. 

  232.         v = v.strip()
    233. 

  233.         self.assertEqual(v, "Leuke/JJ/B-NP/O/O/leuk kat/NN/I-NP/O/O/kat ././O/O/O/.")
    234. 

  234.         print("python -m pattern.nl")
    235. 

  235. 

  236. #---------------------------------------------------------------------------------------------------
    237. 

  237. 

  238. class TestSentiment(unittest.TestCase):
    239. 

  239.     
    240. 

  240.     def setUp(self):
    241. 

  241.         pass
    242. 

  242. 

  243.     def test_sentiment(self):
    244. 

  244.         # Assert < 0 for negative adjectives and > 0 for positive adjectives.
    245. 

  245.         self.assertTrue(nl.sentiment("geweldig")[0] > 0)
    246. 

  246.         self.assertTrue(nl.sentiment("verschrikkelijk")[0] < 0)
    247. 

  247.         # Assert the accuracy of the sentiment analysis.
    248. 

  248.         # Given are the scores for 3,000 book reviews.
    249. 

  249.         # The baseline should increase (not decrease) when the algorithm is modified.
    250. 

  250.         from pattern.db import Datasheet
    251. 

  251.         from pattern.metrics import test
    252. 

  252.         reviews = []
    253. 

  253.         for score, review in Datasheet.load(os.path.join(PATH, "corpora", "polarity-nl-bol.com.csv")):
    254. 

  254.             reviews.append((review, int(score) > 0))
    255. 

  255.         A, P, R, F = test(lambda review: nl.positive(review), reviews)
    256. 

  256.         #print(A, P, R, F)
    257. 

  257.         self.assertTrue(A > 0.808)
    258. 

  258.         self.assertTrue(P > 0.780)
    259. 

  259.         self.assertTrue(R > 0.860)
    260. 

  260.         self.assertTrue(F > 0.818)
    261. 

  261.         print("pattern.nl.sentiment()")
    262. 

  262. 

  263. #---------------------------------------------------------------------------------------------------
    264. 

  264. 

  265. def suite():
    266. 

  266.     suite = unittest.TestSuite()
    267. 

  267.     suite.addTest(unittest.TestLoader().loadTestsFromTestCase(TestInflection))
    268. 

  268.     suite.addTest(unittest.TestLoader().loadTestsFromTestCase(TestParser))
    269. 

  269.     suite.addTest(unittest.TestLoader().loadTestsFromTestCase(TestSentiment))
    270. 

  270.     return suite
    271. 

  271. 

  272. if __name__ == "__main__":
    273. 

  273.     unittest.TextTestRunner(verbosity=1).run(suite())
    274. 

  274.