Datasets:
GEM
/

Tasks:
Other
Multilinguality:
unknown
Size Categories:
unknown
Language Creators:
unknown
Annotations Creators:
expert-created
Source Datasets:
original
License:
Sebastian Gehrmann commited on
Commit
2b5d807
1 Parent(s): f549294

change feature names

Browse files
Files changed (1) hide show
  1. opusparcus.py +36 -14
opusparcus.py CHANGED
@@ -19,6 +19,7 @@
19
  import csv
20
  import json
21
  import os
 
22
  import datasets
23
  import bz2
24
 
@@ -70,6 +71,24 @@ _URLs = {
70
 
71
  _VERSION = datasets.Version("1.0.0", "")
72
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
73
  class OpusparcusConfig(datasets.BuilderConfig):
74
  """BuilderConfig for Opusparcus."""
75
 
@@ -109,7 +128,7 @@ LANGS = [ "de", "en", "fi", "fr", "ru", "sv" ]
109
  # been annotated manually, and each example has an annotation score
110
  # attached to it.)
111
  QUALITIES = [ 100, 95, 90, 85, 80, 75, 70, 65, 60 ]
112
-
113
  class Opusparcus(datasets.GeneratorBasedBuilder):
114
 
115
  """Opusparcus is a paraphrase corpus for six European languages:
@@ -146,12 +165,12 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
146
  # The above commands can alternatively be expressed as:
147
  # data = datasets.load_dataset('GEM/opusparcus', 'de.100')
148
  # data = datasets.load_dataset('GEM/opusparcus', 'fr.75')
149
-
150
  BUILDER_CONFIGS = [
151
  OpusparcusConfig(lang=lang, quality=quality, version=_VERSION) \
152
  for lang in LANGS for quality in QUALITIES
153
  ]
154
-
155
  # There is no default configuration. User always needs to specify one:
156
  # DEFAULT_CONFIG_NAME = None
157
 
@@ -161,10 +180,11 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
161
  features = datasets.Features(
162
  {
163
  "lang": datasets.Value("string"),
164
- "sent1": datasets.Value("string"),
165
- "sent2": datasets.Value("string"),
166
  "annot_score": datasets.Value("float"),
167
  "gem_id": datasets.Value("string"),
 
168
  }
169
  )
170
 
@@ -200,7 +220,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
200
  # This is an error: nothing to do here if no language
201
  # has been defined:
202
  return []
203
-
204
  # Select which file of the training data contains the matching data:
205
  if self.config.quality < 70:
206
  # We need to retrieve the largest training set file
@@ -216,7 +236,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
216
  # download any training data, because there is no matching data.
217
  # The validation and test sets are so small that we do not perform
218
  # any filtering or optimization at this stage.
219
-
220
  # dl_manager is a datasets.download.DownloadManager, which
221
  # downloads and extracts the URLs
222
  # (It can accept any type or nested list/dict and will give
@@ -267,12 +287,12 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
267
  "split": "validation.full",
268
  },
269
  ),
270
- ]
271
 
272
  # If the desired quality value is 100, no subset of the
273
  # training set is good enough, and we only produce validation
274
  # and test sets, in order to save space and time:
275
-
276
  if self.config.quality <= 95:
277
  # In this case there is matching training data, so we produce
278
  # a train split.
@@ -306,7 +326,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
306
  # They contain a field "quality" missing from the validation and test sets.
307
  # We also know that this file only contains the desired language,
308
  # because for the training sets the languages are in separate
309
- # files, and only the desired language has been downloaded.
310
  with bz2.open(filepath, "rt", encoding="utf-8") as f:
311
  for id_, row in enumerate(f):
312
  data = json.loads(row)
@@ -316,10 +336,11 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
316
  break
317
  yield id_, {
318
  "lang": data["lang"],
319
- "sent1": data["sent1"],
320
- "sent2": data["sent2"],
321
  "annot_score": 0.0, # means there is no annotation
322
  "gem_id": data["gem_id"],
 
323
  }
324
  else:
325
  # The validation and test sets are in jsonl files.
@@ -340,9 +361,10 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
340
  # "good or mostly good example of paraphrases")
341
  yield id_, {
342
  "lang": data["lang"],
343
- "sent1": data["sent1"],
344
- "sent2": data["sent2"],
345
  "annot_score": data["annot_score"],
346
  "gem_id": data["gem_id"],
 
347
  }
348
 
19
  import csv
20
  import json
21
  import os
22
+ import re
23
  import datasets
24
  import bz2
25
 
71
 
72
  _VERSION = datasets.Version("1.0.0", "")
73
 
74
+
75
+ def detokenize(text):
76
+ """
77
+ Untokenizing a text undoes the tokenizing operation, restoring
78
+ punctuation and spaces to the places that people expect them to be.
79
+ Ideally, `untokenize(tokenize(text))` should be identical to `text`,
80
+ except for line breaks.
81
+ """
82
+ step1 = text.replace("`` ", '"').replace(" ''", '"').replace('. . .', '...')
83
+ step2 = step1.replace(" ( ", " (").replace(" ) ", ") ")
84
+ step3 = re.sub(r' ([.,:;?!%]+)([ \'"`])', r"\1\2", step2)
85
+ step4 = re.sub(r' ([.,:;?!%]+)$', r"\1", step3)
86
+ step5 = step4.replace(" '", "'").replace(" n't", "n't").replace(
87
+ "can not", "cannot").replace(" 've", "'ve")
88
+ step6 = step5.replace(" ` ", " '")
89
+ return step6.strip()
90
+
91
+
92
  class OpusparcusConfig(datasets.BuilderConfig):
93
  """BuilderConfig for Opusparcus."""
94
 
128
  # been annotated manually, and each example has an annotation score
129
  # attached to it.)
130
  QUALITIES = [ 100, 95, 90, 85, 80, 75, 70, 65, 60 ]
131
+
132
  class Opusparcus(datasets.GeneratorBasedBuilder):
133
 
134
  """Opusparcus is a paraphrase corpus for six European languages:
165
  # The above commands can alternatively be expressed as:
166
  # data = datasets.load_dataset('GEM/opusparcus', 'de.100')
167
  # data = datasets.load_dataset('GEM/opusparcus', 'fr.75')
168
+
169
  BUILDER_CONFIGS = [
170
  OpusparcusConfig(lang=lang, quality=quality, version=_VERSION) \
171
  for lang in LANGS for quality in QUALITIES
172
  ]
173
+
174
  # There is no default configuration. User always needs to specify one:
175
  # DEFAULT_CONFIG_NAME = None
176
 
180
  features = datasets.Features(
181
  {
182
  "lang": datasets.Value("string"),
183
+ "input": datasets.Value("string"),
184
+ "target": datasets.Value("string"),
185
  "annot_score": datasets.Value("float"),
186
  "gem_id": datasets.Value("string"),
187
+ "references": [datasets.Value("string")]
188
  }
189
  )
190
 
220
  # This is an error: nothing to do here if no language
221
  # has been defined:
222
  return []
223
+
224
  # Select which file of the training data contains the matching data:
225
  if self.config.quality < 70:
226
  # We need to retrieve the largest training set file
236
  # download any training data, because there is no matching data.
237
  # The validation and test sets are so small that we do not perform
238
  # any filtering or optimization at this stage.
239
+
240
  # dl_manager is a datasets.download.DownloadManager, which
241
  # downloads and extracts the URLs
242
  # (It can accept any type or nested list/dict and will give
287
  "split": "validation.full",
288
  },
289
  ),
290
+ ]
291
 
292
  # If the desired quality value is 100, no subset of the
293
  # training set is good enough, and we only produce validation
294
  # and test sets, in order to save space and time:
295
+
296
  if self.config.quality <= 95:
297
  # In this case there is matching training data, so we produce
298
  # a train split.
326
  # They contain a field "quality" missing from the validation and test sets.
327
  # We also know that this file only contains the desired language,
328
  # because for the training sets the languages are in separate
329
+ # files, and only the desired language has been downloaded.
330
  with bz2.open(filepath, "rt", encoding="utf-8") as f:
331
  for id_, row in enumerate(f):
332
  data = json.loads(row)
336
  break
337
  yield id_, {
338
  "lang": data["lang"],
339
+ "input": detokenize(data["sent1"]),
340
+ "target": detokenize(data["sent2"]),
341
  "annot_score": 0.0, # means there is no annotation
342
  "gem_id": data["gem_id"],
343
+ "references": [detokenize(data["sent2"])]
344
  }
345
  else:
346
  # The validation and test sets are in jsonl files.
361
  # "good or mostly good example of paraphrases")
362
  yield id_, {
363
  "lang": data["lang"],
364
+ "input": detokenize(data["sent1"]),
365
+ "target": detokenize(data["sent2"]),
366
  "annot_score": data["annot_score"],
367
  "gem_id": data["gem_id"],
368
+ "references": [detokenize(data["sent2"])]
369
  }
370