Upload sumipcc_dataset.py
Browse files- sumipcc_dataset.py +289 -0
sumipcc_dataset.py
ADDED
@@ -0,0 +1,289 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# coding=utf-8
|
2 |
+
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
|
16 |
+
"""Sum-IPCC: Climate Change Report Summarization with Large Language Models"""
|
17 |
+
|
18 |
+
import os
|
19 |
+
import datasets
|
20 |
+
import yaml
|
21 |
+
import textwrap
|
22 |
+
|
23 |
+
|
24 |
+
_CITATION = """
|
25 |
+
@misc{IPCC,
|
26 |
+
title={Climate Change Report Summarization with Large Language Models},
|
27 |
+
author={Leonardo Catalano, Tommaso Colella and Iacopo Ghinassi},
|
28 |
+
year={2024},
|
29 |
+
eprint={},
|
30 |
+
archivePrefix={},
|
31 |
+
primaryClass={}
|
32 |
+
}
|
33 |
+
"""
|
34 |
+
|
35 |
+
_DESCRIPTION = """\
|
36 |
+
A dataset for summarization for policy makers of IPCC synthesis reports for climate change.
|
37 |
+
"""
|
38 |
+
|
39 |
+
_HOMEPAGE = ""
|
40 |
+
|
41 |
+
_LICENSE = ""
|
42 |
+
|
43 |
+
_SumIPCCsum_BASE_KWARGS = dict(
|
44 |
+
citation=_CITATION,
|
45 |
+
url=_HOMEPAGE,
|
46 |
+
)
|
47 |
+
|
48 |
+
|
49 |
+
class SumIPCCConfig(datasets.BuilderConfig):
|
50 |
+
"""BuilderConfig for Climabench."""
|
51 |
+
|
52 |
+
def __init__(
|
53 |
+
self,
|
54 |
+
data_dir,
|
55 |
+
citation,
|
56 |
+
url,
|
57 |
+
**kwargs,
|
58 |
+
):
|
59 |
+
"""BuilderConfig for Climabench.
|
60 |
+
Args:
|
61 |
+
data_dir: `string`, the path to the folder containing the tsv files in the
|
62 |
+
downloaded zip
|
63 |
+
"""
|
64 |
+
super(SumIPCCConfig, self).__init__(
|
65 |
+
version=datasets.Version("1.0.0", ""), **kwargs
|
66 |
+
)
|
67 |
+
self.data_dir = data_dir
|
68 |
+
self.citation = citation
|
69 |
+
self.url = url
|
70 |
+
|
71 |
+
|
72 |
+
class SumIPCC(datasets.GeneratorBasedBuilder):
|
73 |
+
"""SumIPCC dataset."""
|
74 |
+
|
75 |
+
BUILDER_CONFIGS = [
|
76 |
+
SumIPCCConfig(
|
77 |
+
name="AR6",
|
78 |
+
description=textwrap.dedent(
|
79 |
+
"""
|
80 |
+
This Synthesis Report (SYR) of the IPCC Sixth Assessment Report (AR6)
|
81 |
+
summarises the state of knowledge of climate change, its widespread
|
82 |
+
impacts and risks, and climate change mitigation and adaptation, based
|
83 |
+
on the peer-reviewed scientific, technical and socio-economic literature
|
84 |
+
since the publication of the IPCC’s Fifth Assessment Report (AR5) in
|
85 |
+
2014.
|
86 |
+
The assessment is undertaken within the context of the evolving
|
87 |
+
international landscape, in particular, developments in the UN
|
88 |
+
Framework Convention on Climate Change (UNFCCC) process,
|
89 |
+
including the outcomes of the Kyoto Protocol and the adoption of the
|
90 |
+
Paris Agreement. It reflects the increasing diversity of those involved in
|
91 |
+
climate action.
|
92 |
+
This report integrates the main findings of the AR6 Working Group
|
93 |
+
reports58 and the three AR6 Special Reports 59
|
94 |
+
. It recognizes the
|
95 |
+
interdependence of climate, ecosystems and biodiversity, and human
|
96 |
+
societies; the value of diverse forms of knowledge; and the close
|
97 |
+
linkages between climate change adaptation, mitigation, ecosystem
|
98 |
+
health, human well-being and sustainable development. Building on
|
99 |
+
multiple analytical frameworks, including those from the physical and
|
100 |
+
social sciences, this report identifies opportunities for transformative
|
101 |
+
action which are effective, feasible, just and equitable using concepts
|
102 |
+
of systems transitions and resilient development pathways 60
|
103 |
+
. Different
|
104 |
+
regional classification schemes 61 are used for physical, social and
|
105 |
+
economic aspects, reflecting the underlying literature.
|
106 |
+
After this introduction, Section 2, ‘Current Status and Trends’, opens
|
107 |
+
with the assessment of observational evidence for our changing
|
108 |
+
climate, historical and current drivers of human-induced climate
|
109 |
+
change, and its impacts. It assesses the current implementation of
|
110 |
+
adaptation and mitigation response options. Section 3, ‘Long-Term
|
111 |
+
Climate and Development Futures’, provides a long-term assessment of
|
112 |
+
climate change to 2100 and beyond in a broad range of socio-economic
|
113 |
+
futures. It considers long-term characteristics, impacts, risks and costs
|
114 |
+
in adaptation and mitigation pathways in the context of sustainable
|
115 |
+
development. Section 4, ‘Near- Term Responses in a Changing Climate’,
|
116 |
+
assesses opportunities for scaling up effective action in the period up
|
117 |
+
to 2040, in the context of climate pledges, and commitments, and the
|
118 |
+
pursuit of sustainable development.
|
119 |
+
Based on scientific understanding, key findings can be formulated as
|
120 |
+
statements of fact or associated with an assessed level of confidence
|
121 |
+
using the IPCC calibrated language62
|
122 |
+
. The scientific findings are
|
123 |
+
drawn from the underlying reports and arise from their Summary for
|
124 |
+
Policymakers (hereafter SPM), Technical Summary (hereafter TS), and
|
125 |
+
underlying chapters and are indicated by {} brackets. Figure 1.1 shows
|
126 |
+
the Synthesis Report Figures Key, a guide to visual icons that are used
|
127 |
+
across multiple figures within this report.
|
128 |
+
"""
|
129 |
+
),
|
130 |
+
data_dir="all_data/AR6",
|
131 |
+
citation=textwrap.dedent(
|
132 |
+
"""\
|
133 |
+
@misc{IPCCAR6,
|
134 |
+
title = {AR6 Synthesis Report Climate Change 2023},
|
135 |
+
author = {IPCC},
|
136 |
+
year={2024}
|
137 |
+
url = {https://www.ipcc.ch/report/ar6/syr/downloads/report/IPCC_AR6_SYR_LongerReport.pdf},
|
138 |
+
}
|
139 |
+
}"""
|
140 |
+
),
|
141 |
+
url="https://www.ipcc.ch/report/ar6/syr/",
|
142 |
+
),
|
143 |
+
SumIPCCConfig(
|
144 |
+
name="AR5",
|
145 |
+
description=textwrap.dedent(
|
146 |
+
"""
|
147 |
+
The Synthesis Report (SYR), constituting the final product of the Fifth
|
148 |
+
Assessment Report (AR5) of the Intergovernmental Panel on Climate
|
149 |
+
Change (IPCC), is published under the title Climate Change 2014. This
|
150 |
+
report distils, synthesizes and integrates the key findings of the three
|
151 |
+
Working Group contributions – The Physical Science Basis, Impacts,
|
152 |
+
Adaptation, and Vulnerability and Mitigation of Climate Change – to
|
153 |
+
the AR5 in a concise document for the benefit of decision makers in
|
154 |
+
the government, the private sector as well as the public at large. The
|
155 |
+
SYR also draws on the findings of the two Special Reports brought out
|
156 |
+
in 2011 dealing with Renewable Energy Sources and Climate Change
|
157 |
+
Mitigation, and Managing the Risks of Extreme Events and Disasters to
|
158 |
+
Advance Climate Change Adaptation. The SYR, therefore, is a compre-
|
159 |
+
hensive up-to-date compilation of assessments dealing with climate
|
160 |
+
change, based on the most recent scientific, technical and socio-economic
|
161 |
+
literature in the field.
|
162 |
+
"""
|
163 |
+
),
|
164 |
+
data_dir="all_data/AR5",
|
165 |
+
citation=textwrap.dedent(
|
166 |
+
"""\
|
167 |
+
@misc{IPCCAR5,
|
168 |
+
title = {AR5 Synthesis Report: Climate Change 2014},
|
169 |
+
author = {IPCC},
|
170 |
+
year={2014}
|
171 |
+
url = {https://www.ipcc.ch/site/assets/uploads/2018/02/SYR_AR5_FINAL_full.pdf},
|
172 |
+
}
|
173 |
+
}"""
|
174 |
+
),
|
175 |
+
url="https://www.ipcc.ch/report/ar5/syr/",
|
176 |
+
),
|
177 |
+
SumIPCCConfig(
|
178 |
+
name="ALL",
|
179 |
+
description=textwrap.dedent(
|
180 |
+
"""
|
181 |
+
The concatenation of AR5 and AR6 synthesis reports from IPCC
|
182 |
+
"""
|
183 |
+
),
|
184 |
+
data_dir="all_data",
|
185 |
+
citation=textwrap.dedent(
|
186 |
+
"""\
|
187 |
+
@misc{IPCC,
|
188 |
+
title = {IPCC Homepage},
|
189 |
+
author = {IPCC},
|
190 |
+
year={2024}
|
191 |
+
url = {https://www.ipcc.ch/},
|
192 |
+
}
|
193 |
+
}"""
|
194 |
+
),
|
195 |
+
url="https://www.ipcc.ch/"
|
196 |
+
)
|
197 |
+
]
|
198 |
+
|
199 |
+
def _info(self):
|
200 |
+
features = datasets.Features(
|
201 |
+
{
|
202 |
+
"full_paragraphs": datasets.Sequence(datasets.Value("string")),
|
203 |
+
"summary": datasets.Value("string"),
|
204 |
+
"summary_topic": datasets.Value("string"),
|
205 |
+
"paragraph_topic": datasets.Value("string"),
|
206 |
+
"section_topic": datasets.Value("string"),
|
207 |
+
"source": datasets.Value("string"),
|
208 |
+
"paragraph_ids": datasets.Sequence(datasets.Value("string")),
|
209 |
+
"paragraph_titles": datasets.Sequence(datasets.Value("string")),
|
210 |
+
"ID": datasets.Value("string")
|
211 |
+
}
|
212 |
+
)
|
213 |
+
|
214 |
+
return datasets.DatasetInfo(
|
215 |
+
description=_DESCRIPTION,
|
216 |
+
features=datasets.Features(features),
|
217 |
+
homepage=self.config.url,
|
218 |
+
citation=self.config.citation + "\n" + _CITATION,
|
219 |
+
)
|
220 |
+
|
221 |
+
def _split_generators(self, dl_manager):
|
222 |
+
data_dir = self.config.data_dir
|
223 |
+
|
224 |
+
if self.config.name == "ALL":
|
225 |
+
files = []
|
226 |
+
for root, directs, _ in os.walk(data_dir):
|
227 |
+
for direct in directs:
|
228 |
+
file_name = [file for file in os.listdir(os.path.join(root, direct)) if file.endswith("yaml")]
|
229 |
+
assert len(file_name) == 1, "Too many yaml files in directory"
|
230 |
+
file_name = file_name[0]
|
231 |
+
files.append(os.path.join(root, direct, file_name))
|
232 |
+
else:
|
233 |
+
files = []
|
234 |
+
for root, _, fls in os.walk(data_dir):
|
235 |
+
file_name = [file for file in fls if file.endswith("yaml")]
|
236 |
+
assert len(file_name) == 1, "Too many yaml files in directory"
|
237 |
+
file_name = file_name[0]
|
238 |
+
files.append(os.path.join(root, file_name))
|
239 |
+
|
240 |
+
return [
|
241 |
+
datasets.SplitGenerator(
|
242 |
+
name=datasets.Split.TEST,
|
243 |
+
gen_kwargs={
|
244 |
+
"data_files": files,
|
245 |
+
"split": "test",
|
246 |
+
},
|
247 |
+
),
|
248 |
+
]
|
249 |
+
|
250 |
+
def _generate_examples(self, data_files, split):
|
251 |
+
#idx = iter(range(10000))
|
252 |
+
idx = -1
|
253 |
+
for data_file in data_files:
|
254 |
+
with open(data_file, encoding='utf-8') as f:
|
255 |
+
doc = yaml.safe_load(f)
|
256 |
+
for identifier in doc["summaries"]:
|
257 |
+
idx += 1
|
258 |
+
yield from self._process_example(doc, identifier, data_file, idx)
|
259 |
+
|
260 |
+
def _process_example(self, doc,
|
261 |
+
identifier,
|
262 |
+
source,
|
263 |
+
idx):
|
264 |
+
|
265 |
+
summary = doc["summaries"][identifier]
|
266 |
+
full_para = doc["full_paragraphs"][identifier]
|
267 |
+
|
268 |
+
para_topic = doc["paragraph_topics"][identifier]
|
269 |
+
sect_topic = doc["section_topics"][identifier]
|
270 |
+
summary_topic = doc["summary_topics"][identifier]
|
271 |
+
|
272 |
+
para_id = doc["pointers"][identifier]
|
273 |
+
para_title = doc["titles"][identifier]
|
274 |
+
|
275 |
+
yield idx, {"full_paragraphs": full_para,
|
276 |
+
"summary": summary,
|
277 |
+
"paragraph_topic": para_topic,
|
278 |
+
"section_topic": sect_topic,
|
279 |
+
"summary_topic": summary_topic,
|
280 |
+
"source": source,
|
281 |
+
"paragraph_ids": para_id,
|
282 |
+
"paragraph_titles": para_title,
|
283 |
+
"ID": identifier}
|
284 |
+
|
285 |
+
# do this in some postprocess function after generating yaml
|
286 |
+
# def _clean(self, text):
|
287 |
+
# text = re.sub("\n", " ", text)
|
288 |
+
# text = re.sub("'", "", text)
|
289 |
+
# return text.strip()
|