Abinaya Mahendiran
commited on
Commit
•
ce90d54
1
Parent(s):
1a9b60a
Added data loader script - xsum
Browse files
xsum.py
ADDED
@@ -0,0 +1,119 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import os
|
3 |
+
|
4 |
+
import datasets
|
5 |
+
|
6 |
+
_CITATION = """\
|
7 |
+
@article{Narayan2018DontGM,
|
8 |
+
title={Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization},
|
9 |
+
author={Shashi Narayan and Shay B. Cohen and Mirella Lapata},
|
10 |
+
journal={ArXiv},
|
11 |
+
year={2018},
|
12 |
+
volume={abs/1808.08745}
|
13 |
+
}
|
14 |
+
"""
|
15 |
+
|
16 |
+
_DESCRIPTION = """\
|
17 |
+
This is the XSUM subset of the GEM benchmark.
|
18 |
+
"""
|
19 |
+
_URLs = {
|
20 |
+
"xsum": {
|
21 |
+
"data": "http://bollin.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz",
|
22 |
+
"splits": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_xsum_confidence_0.8.json",
|
23 |
+
"challenge_set": "https://storage.googleapis.com/huggingface-nlp/datasets/gem/gem_challenge_sets/xsum.zip",
|
24 |
+
},
|
25 |
+
}
|
26 |
+
|
27 |
+
_XSUM_REMOVE_LINES = set(
|
28 |
+
[
|
29 |
+
"Share this with\n",
|
30 |
+
"Email\n",
|
31 |
+
"Facebook\n",
|
32 |
+
"Messenger\n",
|
33 |
+
"Twitter\n",
|
34 |
+
"Pinterest\n",
|
35 |
+
"WhatsApp\n",
|
36 |
+
"Linkedin\n",
|
37 |
+
"LinkedIn\n",
|
38 |
+
"Copy this link\n",
|
39 |
+
"These are external links and will open in a new window\n",
|
40 |
+
]
|
41 |
+
)
|
42 |
+
|
43 |
+
class Xsum(datasets.GeneratorBasedBuilder):
|
44 |
+
|
45 |
+
BUILDER_CONFIGS = [
|
46 |
+
datasets.BuilderConfig(
|
47 |
+
name=lang,
|
48 |
+
version=datasets.Version("1.0.0"),
|
49 |
+
description="",
|
50 |
+
)
|
51 |
+
]
|
52 |
+
|
53 |
+
def _info(self):
|
54 |
+
return datasets.DatasetInfo(
|
55 |
+
description=_DESCRIPTION,
|
56 |
+
features = datasets.Features(
|
57 |
+
{
|
58 |
+
"gem_id": datasets.Value("string"),
|
59 |
+
"gem_parent_id": datasets.Value("string"),
|
60 |
+
"xsum_id": datasets.Value("string"),
|
61 |
+
"document": datasets.Value("string"),
|
62 |
+
"target": datasets.Value("string"),
|
63 |
+
"references": [datasets.Value("string")],
|
64 |
+
}
|
65 |
+
),
|
66 |
+
supervised_keys=None,
|
67 |
+
homepage="",
|
68 |
+
citation=_CITATION,
|
69 |
+
)
|
70 |
+
|
71 |
+
def _split_generators(self, dl_manager):
|
72 |
+
"""Returns SplitGenerators."""
|
73 |
+
dl_dir = dl_manager.download_and_extract(_URLs[self.config.name])
|
74 |
+
challenge_sets = [
|
75 |
+
("challenge_train_sample", "train_xsum_RandomSample500.json"),
|
76 |
+
("challenge_validation_sample", "validation_xsum_RandomSample500.json"),
|
77 |
+
("challenge_test_backtranslation", "test_xsum_BackTranslation500.json"),
|
78 |
+
("challenge_test_bfp_02", "test_xsum_ButterFingersPerturbation_p=0.02_500.json"),
|
79 |
+
("challenge_test_bfp_05", "test_xsum_ButterFingersPerturbation_p=0.05_500.json"),
|
80 |
+
("challenge_test_nopunc", "test_xsum_WithoutPunctuation500.json"),
|
81 |
+
("challenge_test_covid", f"en_test_covid19.jsonl"),
|
82 |
+
]
|
83 |
+
return [
|
84 |
+
datasets.SplitGenerator(
|
85 |
+
name=challenge_split,
|
86 |
+
gen_kwargs={
|
87 |
+
"filepath": os.path.join(dl_dir["challenge_set"], "xsum", filename),
|
88 |
+
"split": challenge_split,
|
89 |
+
},
|
90 |
+
)
|
91 |
+
for challenge_split, filename in challenge_sets
|
92 |
+
]
|
93 |
+
|
94 |
+
def _generate_examples(self, filepath, split, filepaths=None, lang=None):
|
95 |
+
"""Yields examples."""
|
96 |
+
if "challenge" in split:
|
97 |
+
if "covid" in split:
|
98 |
+
with open(filepath, encoding="utf-8") as f:
|
99 |
+
id_ = -1
|
100 |
+
for line in f:
|
101 |
+
data = json.loads(line)
|
102 |
+
id_ += 1
|
103 |
+
yield id_, {
|
104 |
+
"gem_id": f"{self.config.name}-{split}-{id_}",
|
105 |
+
"gem_parent_id": f"{self.config.name}-{split}-{id_}",
|
106 |
+
"xsum_id": data["url"],
|
107 |
+
"document": data["text"],
|
108 |
+
"target": data["summary"],
|
109 |
+
"references": [] if split == "train" else [data["summary"]],
|
110 |
+
}
|
111 |
+
else:
|
112 |
+
exples = json.load(open(filepath, encoding="utf-8"))
|
113 |
+
if isinstance(exples, dict):
|
114 |
+
assert len(exples) == 1, "multiple entries found"
|
115 |
+
exples = list(exples.values())[0]
|
116 |
+
for id_, exple in enumerate(exples):
|
117 |
+
exple["gem_parent_id"] = exple["gem_id"]
|
118 |
+
exple["gem_id"] = f"{self.config.name}-{split}-{id_}"
|
119 |
+
yield id_, exple
|