Datasets:
ju-resplande
commited on
Commit
·
fb0b3f2
1
Parent(s):
5647b49
Upload plue.py
Browse files
plue.py
CHANGED
@@ -215,7 +215,7 @@ class Plue(datasets.GeneratorBasedBuilder):
|
|
215 |
url="https://www.microsoft.com/en-us/download/details.aspx?id=52398",
|
216 |
),
|
217 |
PlueConfig(
|
218 |
-
name="
|
219 |
description=textwrap.dedent(
|
220 |
"""\
|
221 |
The Quora Question Pairs2 dataset is a collection of question pairs from the
|
@@ -351,6 +351,35 @@ class Plue(datasets.GeneratorBasedBuilder):
|
|
351 |
),
|
352 |
url="https://rajpurkar.github.io/SQuAD-explorer/",
|
353 |
),
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
354 |
PlueConfig(
|
355 |
name="rte",
|
356 |
description=textwrap.dedent(
|
|
|
215 |
url="https://www.microsoft.com/en-us/download/details.aspx?id=52398",
|
216 |
),
|
217 |
PlueConfig(
|
218 |
+
name="qqp_v2",
|
219 |
description=textwrap.dedent(
|
220 |
"""\
|
221 |
The Quora Question Pairs2 dataset is a collection of question pairs from the
|
|
|
351 |
),
|
352 |
url="https://rajpurkar.github.io/SQuAD-explorer/",
|
353 |
),
|
354 |
+
PlueConfig(
|
355 |
+
name="qnli_v2",
|
356 |
+
description=textwrap.dedent(
|
357 |
+
"""\
|
358 |
+
The Stanford Question Answering Dataset is a question-answering
|
359 |
+
dataset consisting of question-paragraph pairs, where one of the sentences in the paragraph (drawn
|
360 |
+
from Wikipedia) contains the answer to the corresponding question (written by an annotator). We
|
361 |
+
convert the task into sentence pair classification by forming a pair between each question and each
|
362 |
+
sentence in the corresponding context, and filtering out pairs with low lexical overlap between the
|
363 |
+
question and the context sentence. The task is to determine whether the context sentence contains
|
364 |
+
the answer to the question. This modified version of the original task removes the requirement that
|
365 |
+
the model select the exact answer, but also removes the simplifying assumptions that the answer
|
366 |
+
is always present in the input and that lexical overlap is a reliable cue."""
|
367 |
+
), # pylint: disable=line-too-long
|
368 |
+
text_features={"question": "question", "sentence": "sentence",},
|
369 |
+
label_classes=["entailment", "not_entailment"],
|
370 |
+
label_column="label",
|
371 |
+
data_dir="PLUE-1.0.1/datasets/QNLI_v2",
|
372 |
+
citation=textwrap.dedent(
|
373 |
+
"""\
|
374 |
+
@article{rajpurkar2016squad,
|
375 |
+
title={Squad: 100,000+ questions for machine comprehension of text},
|
376 |
+
author={Rajpurkar, Pranav and Zhang, Jian and Lopyrev, Konstantin and Liang, Percy},
|
377 |
+
journal={arXiv preprint arXiv:1606.05250},
|
378 |
+
year={2016}
|
379 |
+
}"""
|
380 |
+
),
|
381 |
+
url="https://rajpurkar.github.io/SQuAD-explorer/",
|
382 |
+
),
|
383 |
PlueConfig(
|
384 |
name="rte",
|
385 |
description=textwrap.dedent(
|