Datasets:
GEM
/

Modalities:
Text
Languages:
English
ArXiv:
Libraries:
Datasets
License:
Sebastian Gehrmann commited on
Commit
2b42c85
1 Parent(s): ee678f8
Files changed (1) hide show
  1. wiki_cat_sum.py +84 -83
wiki_cat_sum.py CHANGED
@@ -51,71 +51,71 @@ _LICENSE = "CC BY-SA 3.0"
51
  # This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
52
  _URLs = {
53
  "animal": {
54
- "train": "./main_splits/train-animal.jsonl",
55
- "validation": "./main_splits/valid-animal.jsonl",
56
- "test": "./main_splits/test-animal.jsonl",
57
- "cs_abs":[
58
- "./cs_abs/test-animal_nv_0.jsonl",
59
- "./cs_abs/test-animal_nv_1.jsonl",
60
- "./cs_abs/test-animal_nv_2.jsonl",
61
- "./cs_abs/test-animal_nv_3.jsonl",
62
- "./cs_abs/test-animal_nv_4.jsonl",
63
- "./cs_abs/test-animal_nv_6.jsonl",
64
- "./cs_abs/test-animal_nv_7.jsonl",
65
- "./cs_abs/test-animal_nv_8.jsonl",
66
- "./cs_abs/test-animal_nv_9.jsonl",
67
  ],
68
  "cs_tdiv": [
69
- "./cs_tdiv/test-animal_tdiv_0.jsonl",
70
- "./cs_tdiv/test-animal_tdiv_1.jsonl",
71
- "./cs_tdiv/test-animal_tdiv_2.jsonl",
72
- "./cs_tdiv/test-animal_tdiv_3.jsonl",
73
- ]
74
  },
75
  "company": {
76
- "train": "./main_splits/train-company.jsonl",
77
- "validation": "./main_splits/valid-company.jsonl",
78
- "test": "./main_splits/test-company.jsonl",
79
- "cs_abs":[
80
- "./cs_abs/test-company_nv_0.jsonl",
81
- "./cs_abs/test-company_nv_1.jsonl",
82
- "./cs_abs/test-company_nv_2.jsonl",
83
- "./cs_abs/test-company_nv_3.jsonl",
84
- "./cs_abs/test-company_nv_4.jsonl",
85
- "./cs_abs/test-company_nv_6.jsonl",
86
- "./cs_abs/test-company_nv_7.jsonl",
87
- "./cs_abs/test-company_nv_8.jsonl",
88
- "./cs_abs/test-company_nv_9.jsonl",
89
  ],
90
  "cs_tdiv": [
91
- "./cs_tdiv/test-company_tdiv_0.jsonl",
92
- "./cs_tdiv/test-company_tdiv_1.jsonl",
93
- "./cs_tdiv/test-company_tdiv_2.jsonl",
94
- "./cs_tdiv/test-company_tdiv_3.jsonl",
95
- ]
96
  },
97
  "film": {
98
- "train": "./main_splits/train-film.jsonl",
99
- "validation": "./main_splits/valid-film.jsonl",
100
- "test": "./main_splits/test-film.jsonl",
101
- "cs_abs":[
102
- "./cs_abs/test-film_nv_0.jsonl",
103
- "./cs_abs/test-film_nv_1.jsonl",
104
- "./cs_abs/test-film_nv_2.jsonl",
105
- "./cs_abs/test-film_nv_3.jsonl",
106
- "./cs_abs/test-film_nv_4.jsonl",
107
- "./cs_abs/test-film_nv_6.jsonl",
108
- "./cs_abs/test-film_nv_7.jsonl",
109
- "./cs_abs/test-film_nv_8.jsonl",
110
- "./cs_abs/test-film_nv_9.jsonl",
111
  ],
112
  "cs_tdiv": [
113
- "./cs_tdiv/test-film_tdiv_0.jsonl",
114
- "./cs_tdiv/test-film_tdiv_1.jsonl",
115
- "./cs_tdiv/test-film_tdiv_2.jsonl",
116
- "./cs_tdiv/test-film_tdiv_3.jsonl",
117
- ]
118
- }
119
  }
120
 
121
 
@@ -125,7 +125,6 @@ class WikiCatSum(datasets.GeneratorBasedBuilder):
125
 
126
  VERSION = datasets.Version("0.1.0")
127
 
128
-
129
  # If you need to make complex sub-parts in the datasets with configurable options
130
  # You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig
131
  # BUILDER_CONFIG_CLASS = MyBuilderConfig
@@ -134,9 +133,13 @@ class WikiCatSum(datasets.GeneratorBasedBuilder):
134
  # data = datasets.load_dataset('my_dataset', 'first_domain')
135
  # data = datasets.load_dataset('my_dataset', 'second_domain')
136
  BUILDER_CONFIGS = [
137
- datasets.BuilderConfig(name="animal" , version=VERSION, description="Animal domain"),
138
- datasets.BuilderConfig(name="company", version=VERSION, description="Company domain"),
139
- datasets.BuilderConfig(name="film" , version=VERSION, description="Film domain"),
 
 
 
 
140
  ]
141
 
142
  DEFAULT_CONFIG_NAME = "animal" # It's not mandatory to have a default configuration. Just use one if it make sense.
@@ -149,13 +152,13 @@ class WikiCatSum(datasets.GeneratorBasedBuilder):
149
  "gem_parent_id": datasets.Value("string"),
150
  "id": datasets.Value("string"),
151
  "title": datasets.Value("string"),
152
- "paragraphs": datasets.features.Sequence(
153
- datasets.Value("string")),
154
  "summary": datasets.features.Sequence(
155
  {
156
  "text": datasets.Value("string"),
157
  "topic": datasets.Value("int16"),
158
- })
 
159
  # These are the features of your dataset like images, labels ...
160
  }
161
  )
@@ -187,13 +190,12 @@ class WikiCatSum(datasets.GeneratorBasedBuilder):
187
  my_urls = _URLs[self.config.name]
188
  d_conf = dl_manager.download_and_extract(my_urls)
189
  challenge_sets = [
190
- ("challenge_test_abstractivity_%d" % (lvl), fname) \
191
- for lvl,fname in enumerate(d_conf["cs_abs"])
192
  ] + [
193
- ("challenge_test_topic_diversity_%d" % (lvl), fname) \
194
- for lvl,fname in enumerate(d_conf["cs_abs"])
195
  ]
196
-
197
 
198
  return [
199
  datasets.SplitGenerator(
@@ -207,10 +209,7 @@ class WikiCatSum(datasets.GeneratorBasedBuilder):
207
  datasets.SplitGenerator(
208
  name=datasets.Split.TEST,
209
  # These kwargs will be passed to _generate_examples
210
- gen_kwargs={
211
- "filepath": d_conf["validation"],
212
- "split": "test"
213
- },
214
  ),
215
  datasets.SplitGenerator(
216
  name=datasets.Split.VALIDATION,
@@ -221,20 +220,22 @@ class WikiCatSum(datasets.GeneratorBasedBuilder):
221
  },
222
  ),
223
  ] + [
224
- datasets.SplitGenerator(
225
- name=challenge_split,
226
- gen_kwargs={
227
- "filepath": filename,
228
- "split": challenge_split,
229
- },
230
- )
231
- for challenge_split, filename in challenge_sets
232
- ]
233
 
234
  def _generate_examples(
235
- self, filepath, split # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
 
 
236
  ):
237
- """ Yields examples as (key, example) tuples. """
238
  # This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
239
  # The `key` is here for legacy reason (tfds) and is not important in itself.
240
 
@@ -245,4 +246,4 @@ class WikiCatSum(datasets.GeneratorBasedBuilder):
245
  # data["gem_id"] = "GEM-wiki_cat_sum-%s-%d" % (split,data["id"]+1)
246
  data["gem_parent_id"] = f"{self.config.name}-{split}-{id_+1}"
247
  data["gem_id"] = f"{self.config.name}-{split}-{id_+1}"
248
- yield id_,data
 
51
  # This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
52
  _URLs = {
53
  "animal": {
54
+ "train": "main_splits/train-animal.jsonl",
55
+ "validation": "main_splits/valid-animal.jsonl",
56
+ "test": "main_splits/test-animal.jsonl",
57
+ "cs_abs": [
58
+ "cs_abs/test-animal_nv_0.jsonl",
59
+ "cs_abs/test-animal_nv_1.jsonl",
60
+ "cs_abs/test-animal_nv_2.jsonl",
61
+ "cs_abs/test-animal_nv_3.jsonl",
62
+ "cs_abs/test-animal_nv_4.jsonl",
63
+ "cs_abs/test-animal_nv_6.jsonl",
64
+ "cs_abs/test-animal_nv_7.jsonl",
65
+ "cs_abs/test-animal_nv_8.jsonl",
66
+ "cs_abs/test-animal_nv_9.jsonl",
67
  ],
68
  "cs_tdiv": [
69
+ "cs_tdiv/test-animal_tdiv_0.jsonl",
70
+ "cs_tdiv/test-animal_tdiv_1.jsonl",
71
+ "cs_tdiv/test-animal_tdiv_2.jsonl",
72
+ "cs_tdiv/test-animal_tdiv_3.jsonl",
73
+ ],
74
  },
75
  "company": {
76
+ "train": "main_splits/train-company.jsonl",
77
+ "validation": "main_splits/valid-company.jsonl",
78
+ "test": "main_splits/test-company.jsonl",
79
+ "cs_abs": [
80
+ "cs_abs/test-company_nv_0.jsonl",
81
+ "cs_abs/test-company_nv_1.jsonl",
82
+ "cs_abs/test-company_nv_2.jsonl",
83
+ "cs_abs/test-company_nv_3.jsonl",
84
+ "cs_abs/test-company_nv_4.jsonl",
85
+ "cs_abs/test-company_nv_6.jsonl",
86
+ "cs_abs/test-company_nv_7.jsonl",
87
+ "cs_abs/test-company_nv_8.jsonl",
88
+ "cs_abs/test-company_nv_9.jsonl",
89
  ],
90
  "cs_tdiv": [
91
+ "cs_tdiv/test-company_tdiv_0.jsonl",
92
+ "cs_tdiv/test-company_tdiv_1.jsonl",
93
+ "cs_tdiv/test-company_tdiv_2.jsonl",
94
+ "cs_tdiv/test-company_tdiv_3.jsonl",
95
+ ],
96
  },
97
  "film": {
98
+ "train": "main_splits/train-film.jsonl",
99
+ "validation": "main_splits/valid-film.jsonl",
100
+ "test": "main_splits/test-film.jsonl",
101
+ "cs_abs": [
102
+ "cs_abs/test-film_nv_0.jsonl",
103
+ "cs_abs/test-film_nv_1.jsonl",
104
+ "cs_abs/test-film_nv_2.jsonl",
105
+ "cs_abs/test-film_nv_3.jsonl",
106
+ "cs_abs/test-film_nv_4.jsonl",
107
+ "cs_abs/test-film_nv_6.jsonl",
108
+ "cs_abs/test-film_nv_7.jsonl",
109
+ "cs_abs/test-film_nv_8.jsonl",
110
+ "cs_abs/test-film_nv_9.jsonl",
111
  ],
112
  "cs_tdiv": [
113
+ "cs_tdiv/test-film_tdiv_0.jsonl",
114
+ "cs_tdiv/test-film_tdiv_1.jsonl",
115
+ "cs_tdiv/test-film_tdiv_2.jsonl",
116
+ "cs_tdiv/test-film_tdiv_3.jsonl",
117
+ ],
118
+ },
119
  }
120
 
121
 
 
125
 
126
  VERSION = datasets.Version("0.1.0")
127
 
 
128
  # If you need to make complex sub-parts in the datasets with configurable options
129
  # You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig
130
  # BUILDER_CONFIG_CLASS = MyBuilderConfig
 
133
  # data = datasets.load_dataset('my_dataset', 'first_domain')
134
  # data = datasets.load_dataset('my_dataset', 'second_domain')
135
  BUILDER_CONFIGS = [
136
+ datasets.BuilderConfig(
137
+ name="animal", version=VERSION, description="Animal domain"
138
+ ),
139
+ datasets.BuilderConfig(
140
+ name="company", version=VERSION, description="Company domain"
141
+ ),
142
+ datasets.BuilderConfig(name="film", version=VERSION, description="Film domain"),
143
  ]
144
 
145
  DEFAULT_CONFIG_NAME = "animal" # It's not mandatory to have a default configuration. Just use one if it make sense.
 
152
  "gem_parent_id": datasets.Value("string"),
153
  "id": datasets.Value("string"),
154
  "title": datasets.Value("string"),
155
+ "paragraphs": datasets.features.Sequence(datasets.Value("string")),
 
156
  "summary": datasets.features.Sequence(
157
  {
158
  "text": datasets.Value("string"),
159
  "topic": datasets.Value("int16"),
160
+ }
161
+ )
162
  # These are the features of your dataset like images, labels ...
163
  }
164
  )
 
190
  my_urls = _URLs[self.config.name]
191
  d_conf = dl_manager.download_and_extract(my_urls)
192
  challenge_sets = [
193
+ ("challenge_test_abstractivity_%d" % (lvl), fname)
194
+ for lvl, fname in enumerate(d_conf["cs_abs"])
195
  ] + [
196
+ ("challenge_test_topic_diversity_%d" % (lvl), fname)
197
+ for lvl, fname in enumerate(d_conf["cs_abs"])
198
  ]
 
199
 
200
  return [
201
  datasets.SplitGenerator(
 
209
  datasets.SplitGenerator(
210
  name=datasets.Split.TEST,
211
  # These kwargs will be passed to _generate_examples
212
+ gen_kwargs={"filepath": d_conf["validation"], "split": "test"},
 
 
 
213
  ),
214
  datasets.SplitGenerator(
215
  name=datasets.Split.VALIDATION,
 
220
  },
221
  ),
222
  ] + [
223
+ datasets.SplitGenerator(
224
+ name=challenge_split,
225
+ gen_kwargs={
226
+ "filepath": filename,
227
+ "split": challenge_split,
228
+ },
229
+ )
230
+ for challenge_split, filename in challenge_sets
231
+ ]
232
 
233
  def _generate_examples(
234
+ self,
235
+ filepath,
236
+ split, # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
237
  ):
238
+ """Yields examples as (key, example) tuples."""
239
  # This method handles input defined in _split_generators to yield (key, example) tuples from the dataset.
240
  # The `key` is here for legacy reason (tfds) and is not important in itself.
241
 
 
246
  # data["gem_id"] = "GEM-wiki_cat_sum-%s-%d" % (split,data["id"]+1)
247
  data["gem_parent_id"] = f"{self.config.name}-{split}-{id_+1}"
248
  data["gem_id"] = f"{self.config.name}-{split}-{id_+1}"
249
+ yield id_, data