Felix
commited on
Commit
·
b47191a
1
Parent(s):
6c3740d
add winogender and winograd
Browse files- superlim-2.py +62 -5
superlim-2.py
CHANGED
@@ -94,11 +94,11 @@ _SweSat_CITATION = """\
|
|
94 |
_SweSim_DESCRIPTION = """\
|
95 |
SuperSim is a large-scale similarity and relatedness test set for Swedish built with expert human judgments. The test set is composed of 1360 word-pairs independently judged for both relatedness and similarity by five annotators."""
|
96 |
|
97 |
-
|
98 |
The SweWinogender test set is diagnostic dataset to measure gender bias in coreference resolution. It is modelled after the English Winogender benchmark,
|
99 |
and is released with reference statistics on the distribution of men and women between occupations and the association between gender and occupation in modern corpus material."""
|
100 |
|
101 |
-
|
102 |
SweWinograd is a pronoun resolution test set, containing constructed items in the style of Winograd schema’s. The interpretation of the target pronouns is determined by (common sense)
|
103 |
reasoning and knowledge, and not by syntactic constraints, lexical distributional information or discourse structuring patterns.
|
104 |
The dataset contains 90 multiple choice with multiple correct answers test items."""
|
@@ -129,7 +129,10 @@ _TASKS = {
|
|
129 |
"swesat": "swesat-synonyms",
|
130 |
"swesim_relatedness": "supersim-superlim-relatedness",
|
131 |
"swesim_similarity": "supersim-superlim-similarity",
|
132 |
-
"swewic": "swewic"
|
|
|
|
|
|
|
133 |
}
|
134 |
|
135 |
class SuperLimConfig(datasets.BuilderConfig):
|
@@ -324,6 +327,43 @@ class SuperLim(datasets.GeneratorBasedBuilder):
|
|
324 |
"pos": datasets.Value("string")
|
325 |
})
|
326 |
})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
327 |
else:
|
328 |
raise ValueError(f"Subset {self.config.name} does not exist.")
|
329 |
return datasets.DatasetInfo(
|
@@ -471,12 +511,29 @@ class SuperLim(datasets.GeneratorBasedBuilder):
|
|
471 |
"label": row["label"],
|
472 |
}
|
473 |
elif self.config.name == "swewic":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
474 |
yield key, {
|
475 |
"idx": row["idx"],
|
476 |
-
"
|
477 |
"second": row["second"],
|
478 |
"label": row["label"],
|
479 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
480 |
}
|
481 |
else:
|
482 |
raise ValueError(f"Subset {self.config.name} does not exist")
|
|
|
94 |
_SweSim_DESCRIPTION = """\
|
95 |
SuperSim is a large-scale similarity and relatedness test set for Swedish built with expert human judgments. The test set is composed of 1360 word-pairs independently judged for both relatedness and similarity by five annotators."""
|
96 |
|
97 |
+
_SweWinogender_DESCRIPTION = """\
|
98 |
The SweWinogender test set is diagnostic dataset to measure gender bias in coreference resolution. It is modelled after the English Winogender benchmark,
|
99 |
and is released with reference statistics on the distribution of men and women between occupations and the association between gender and occupation in modern corpus material."""
|
100 |
|
101 |
+
_SweWinograd_DESCRIPTION = """\
|
102 |
SweWinograd is a pronoun resolution test set, containing constructed items in the style of Winograd schema’s. The interpretation of the target pronouns is determined by (common sense)
|
103 |
reasoning and knowledge, and not by syntactic constraints, lexical distributional information or discourse structuring patterns.
|
104 |
The dataset contains 90 multiple choice with multiple correct answers test items."""
|
|
|
129 |
"swesat": "swesat-synonyms",
|
130 |
"swesim_relatedness": "supersim-superlim-relatedness",
|
131 |
"swesim_similarity": "supersim-superlim-similarity",
|
132 |
+
"swewic": "swewic",
|
133 |
+
"swewinogender": "swewinogender",
|
134 |
+
"swewinograd": "swewinograd"
|
135 |
+
|
136 |
}
|
137 |
|
138 |
class SuperLimConfig(datasets.BuilderConfig):
|
|
|
327 |
"pos": datasets.Value("string")
|
328 |
})
|
329 |
})
|
330 |
+
elif self.config.name == 'swewinogender':
|
331 |
+
features = datasets.Features({
|
332 |
+
"idx": datasets.Value(dtype='int32'),
|
333 |
+
'premise': datasets.Value("string"),
|
334 |
+
'hypothesis': datasets.Value("string"),
|
335 |
+
'label': datasets.ClassLabel(num_classes=3, names=['entailment', 'contradiction', 'neutral']),
|
336 |
+
'meta': datasets.Features({
|
337 |
+
'tuple_id': datasets.Value("string"),
|
338 |
+
'template_id': datasets.Value("string"),
|
339 |
+
'occupation_participant': datasets.Value("string"),
|
340 |
+
'other_participant': datasets.Value("string"),
|
341 |
+
'pronoun': datasets.Value("string")
|
342 |
+
})
|
343 |
+
})
|
344 |
+
elif self.config.name == 'swewinograd':
|
345 |
+
features = datasets.Features({
|
346 |
+
"idx": datasets.Value(dtype='int32'),
|
347 |
+
'text': datasets.Value("string"),
|
348 |
+
'label': datasets.ClassLabel(num_classes=2, names=['not_coreferring', 'coreferring']),
|
349 |
+
'pronoun': datasets.Features({
|
350 |
+
'text': datasets.Value("string"),
|
351 |
+
'location': datasets.Features({
|
352 |
+
"start": datasets.Value(dtype='int32'),
|
353 |
+
"stop": datasets.Value(dtype='int32')
|
354 |
+
})
|
355 |
+
}),
|
356 |
+
'candidate_antecedent': datasets.Features({
|
357 |
+
"text": datasets.Value("string"),
|
358 |
+
'location': datasets.Features({
|
359 |
+
"start": datasets.Value(dtype='int32'),
|
360 |
+
"stop": datasets.Value(dtype='int32')
|
361 |
+
})
|
362 |
+
}),
|
363 |
+
'meta': datasets.Features({
|
364 |
+
'snippet_id': datasets.Value("string")
|
365 |
+
})
|
366 |
+
})
|
367 |
else:
|
368 |
raise ValueError(f"Subset {self.config.name} does not exist.")
|
369 |
return datasets.DatasetInfo(
|
|
|
511 |
"label": row["label"],
|
512 |
}
|
513 |
elif self.config.name == "swewic":
|
514 |
+
yield key, {
|
515 |
+
"premise": row["premise"],
|
516 |
+
"hypothesis": row["hypothesis"],
|
517 |
+
"label": row["label"],
|
518 |
+
"label": row["label"],
|
519 |
+
"meta": row["meta"],
|
520 |
+
}
|
521 |
+
elif self.config.name == "swewinogender":
|
522 |
yield key, {
|
523 |
"idx": row["idx"],
|
524 |
+
"text": row["text"],
|
525 |
"second": row["second"],
|
526 |
"label": row["label"],
|
527 |
+
"pronoun": row["pronoun"],
|
528 |
+
}
|
529 |
+
elif self.config.name == "swewinograd":
|
530 |
+
yield key, {
|
531 |
+
"idx": row["idx"],
|
532 |
+
"text": row["text"],
|
533 |
+
"label": row["label"],
|
534 |
+
"pronoun": row["pronoun"],
|
535 |
+
"candidate_antecedent": row["candidate_antecedent"],
|
536 |
+
"meta": row["meta"]
|
537 |
}
|
538 |
else:
|
539 |
raise ValueError(f"Subset {self.config.name} does not exist")
|