Chester Palen-Michel
commited on
Commit
•
563b28c
1
Parent(s):
0c18bf0
Add loading script
Browse files
lr-sum.py
ADDED
@@ -0,0 +1,160 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""LR-Sum summarization dataset"""
|
2 |
+
|
3 |
+
import json
|
4 |
+
import os
|
5 |
+
|
6 |
+
import datasets
|
7 |
+
|
8 |
+
_CITATION = """\
|
9 |
+
@inproceedings{palen-michel-lignos-2023-lr,
|
10 |
+
title = "{LR}-Sum: Summarization for Less-Resourced Languages",
|
11 |
+
author = "Palen-Michel, Chester and
|
12 |
+
Lignos, Constantine",
|
13 |
+
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
|
14 |
+
month = jul,
|
15 |
+
year = "2023",
|
16 |
+
address = "Toronto, Canada",
|
17 |
+
publisher = "Association for Computational Linguistics",
|
18 |
+
url = "https://aclanthology.org/2023.findings-acl.427",
|
19 |
+
doi = "10.18653/v1/2023.findings-acl.427",
|
20 |
+
pages = "6829--6844",
|
21 |
+
abstract = "We introduce LR-Sum, a new permissively-licensed dataset created with the goal of enabling further research in automatic summarization for less-resourced languages.LR-Sum contains human-written summaries for 40 languages, many of which are less-resourced. We describe our process for extracting and filtering the dataset from the Multilingual Open Text corpus (Palen-Michel et al., 2022).The source data is public domain newswire collected from from Voice of America websites, and LR-Sum is released under a Creative Commons license (CC BY 4.0), making it one of the most openly-licensed multilingual summarization datasets. We describe abstractive and extractive summarization experiments to establish baselines and discuss the limitations of this dataset.",
|
22 |
+
}
|
23 |
+
"""
|
24 |
+
|
25 |
+
_DESCRIPTION = """\
|
26 |
+
We introduce LR-Sum, a new permissively-licensed dataset created with the goal of enabling further research in automatic summarization for less-resourced languages.
|
27 |
+
LR-Sum contains human-written summaries for 40 languages, many of which are less-resourced.
|
28 |
+
We describe our process for extracting and filtering the dataset from the Multilingual Open Text corpus (Palen-Michel et al., 2022).
|
29 |
+
The source data is public domain newswire collected from from Voice of America websites, and LR-Sum is released under a Creative Commons license (CC BY 4.0), making it one of the most openly-licensed multilingual summarization datasets.
|
30 |
+
We describe abstractive and extractive summarization experiments to establish baselines and discuss the limitations of this dataset.
|
31 |
+
"""
|
32 |
+
|
33 |
+
_HOMEPAGE = "https://github.com/bltlab"
|
34 |
+
|
35 |
+
_LICENSE = "Creative Commons Attribution 4.0 International (CC-BY 4.0)"
|
36 |
+
|
37 |
+
_URL = "https://huggingface.co/datasets/bltlab/lr-sum/resolve/main/data/{}.zip"
|
38 |
+
|
39 |
+
_LANGUAGES = [
|
40 |
+
"amh",
|
41 |
+
"aze",
|
42 |
+
"ben",
|
43 |
+
"bod",
|
44 |
+
"bos",
|
45 |
+
"ckb",
|
46 |
+
"cmn_t",
|
47 |
+
"cmn_s",
|
48 |
+
"ell",
|
49 |
+
"eng",
|
50 |
+
"fas",
|
51 |
+
"fra",
|
52 |
+
"hat",
|
53 |
+
"hau",
|
54 |
+
"hye",
|
55 |
+
"ind",
|
56 |
+
"kat",
|
57 |
+
"khm",
|
58 |
+
"kin",
|
59 |
+
"kor",
|
60 |
+
"kmr",
|
61 |
+
"lao",
|
62 |
+
"mkd",
|
63 |
+
"mya",
|
64 |
+
"nde",
|
65 |
+
"por",
|
66 |
+
"prs",
|
67 |
+
"pus",
|
68 |
+
"rus",
|
69 |
+
"sna",
|
70 |
+
"som",
|
71 |
+
"spa",
|
72 |
+
"sqi",
|
73 |
+
"srp",
|
74 |
+
"swh",
|
75 |
+
"tha",
|
76 |
+
"tir",
|
77 |
+
"tur",
|
78 |
+
"ukr",
|
79 |
+
"urd",
|
80 |
+
"uzb",
|
81 |
+
"vie",
|
82 |
+
]
|
83 |
+
|
84 |
+
|
85 |
+
class Lrsum(datasets.GeneratorBasedBuilder):
|
86 |
+
VERSION = datasets.Version("1.0.0")
|
87 |
+
|
88 |
+
BUILDER_CONFIGS = [
|
89 |
+
datasets.BuilderConfig(
|
90 |
+
name="{}".format(lang),
|
91 |
+
version=datasets.Version("1.0.0")
|
92 |
+
)
|
93 |
+
for lang in _LANGUAGES
|
94 |
+
]
|
95 |
+
|
96 |
+
def _info(self):
|
97 |
+
return datasets.DatasetInfo(
|
98 |
+
description=_DESCRIPTION,
|
99 |
+
features=datasets.Features(
|
100 |
+
{
|
101 |
+
"id": datasets.Value("string"),
|
102 |
+
"url": datasets.Value("string"),
|
103 |
+
"title": datasets.Value("string"),
|
104 |
+
"summary": datasets.Value("string"),
|
105 |
+
"text": datasets.Value("string"),
|
106 |
+
}
|
107 |
+
),
|
108 |
+
supervised_keys=None,
|
109 |
+
homepage=_HOMEPAGE,
|
110 |
+
citation=_CITATION,
|
111 |
+
license=_LICENSE,
|
112 |
+
version=self.VERSION,
|
113 |
+
)
|
114 |
+
|
115 |
+
def _split_generators(self, dl_manager):
|
116 |
+
"""Returns SplitGenerators."""
|
117 |
+
lang = str(self.config.name)
|
118 |
+
url = _URL.format(lang)
|
119 |
+
|
120 |
+
data_dir = dl_manager.download_and_extract(url)
|
121 |
+
ret = [
|
122 |
+
datasets.SplitGenerator(
|
123 |
+
name=datasets.Split.TEST,
|
124 |
+
gen_kwargs={
|
125 |
+
"filepath": os.path.join(data_dir, lang + "_test.jsonl"),
|
126 |
+
},
|
127 |
+
)
|
128 |
+
]
|
129 |
+
if os.path.exists(os.path.join(data_dir, lang + "_train.jsonl")):
|
130 |
+
ret.append(datasets.SplitGenerator(
|
131 |
+
name=datasets.Split.TRAIN,
|
132 |
+
gen_kwargs={
|
133 |
+
"filepath": os.path.join(data_dir, lang + "_train.jsonl"),
|
134 |
+
},
|
135 |
+
)
|
136 |
+
)
|
137 |
+
if os.path.exists(os.path.join(data_dir, lang + "_val.jsonl")):
|
138 |
+
ret.append(
|
139 |
+
datasets.SplitGenerator(
|
140 |
+
name=datasets.Split.VALIDATION,
|
141 |
+
gen_kwargs={
|
142 |
+
"filepath": os.path.join(data_dir, lang + "_val.jsonl"),
|
143 |
+
},
|
144 |
+
)
|
145 |
+
)
|
146 |
+
|
147 |
+
return ret
|
148 |
+
|
149 |
+
def _generate_examples(self, filepath):
|
150 |
+
"""Yields examples as (key, example) tuples."""
|
151 |
+
with open(filepath, encoding="utf-8") as f:
|
152 |
+
for idx_, row in enumerate(f):
|
153 |
+
data = json.loads(row)
|
154 |
+
yield idx_, {
|
155 |
+
"id": data["id"],
|
156 |
+
"url": data["url"],
|
157 |
+
"title": data["title"],
|
158 |
+
"summary": data["summary"],
|
159 |
+
"text": data["text"],
|
160 |
+
}
|