Datasets:

Modalities:
Text
Languages:
English
Libraries:
Datasets
License:
shuyangcao commited on
Commit
65e803e
1 Parent(s): 545c018

First version

Browse files
.gitattributes CHANGED
@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
35
  *.mp3 filter=lfs diff=lfs merge=lfs -text
36
  *.ogg filter=lfs diff=lfs merge=lfs -text
37
  *.wav filter=lfs diff=lfs merge=lfs -text
 
 
35
  *.mp3 filter=lfs diff=lfs merge=lfs -text
36
  *.ogg filter=lfs diff=lfs merge=lfs -text
37
  *.wav filter=lfs diff=lfs merge=lfs -text
38
+ *.jsonl filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ annotations_creators:
3
+ - no-annotation
4
+ language_creators:
5
+ - expert-generated
6
+ languages:
7
+ - en
8
+ licenses:
9
+ - cc-by-4.0
10
+ multilinguality:
11
+ - monolingual
12
+ size_categories:
13
+ - 10K<n<100K
14
+ source_datasets:
15
+ - original
16
+ task_categories:
17
+ - summarization
18
+ task_ids:
19
+ - summarization
20
+ pretty_name: GovReport
21
+ ---
22
+
23
+
24
+ # Dataset Card for GovReport
25
+
26
+ ## Table of Contents
27
+ - [Table of Contents](#table-of-contents)
28
+ - [Dataset Description](#dataset-description)
29
+ - [Dataset Summary](#dataset-summary)
30
+ - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
31
+ - [Languages](#languages)
32
+ - [Dataset Structure](#dataset-structure)
33
+ - [Data Instances](#data-instances)
34
+ - [Data Fields](#data-fields)
35
+ - [Data Splits](#data-splits)
36
+ - [Dataset Creation](#dataset-creation)
37
+ - [Curation Rationale](#curation-rationale)
38
+ - [Source Data](#source-data)
39
+ - [Annotations](#annotations)
40
+ - [Personal and Sensitive Information](#personal-and-sensitive-information)
41
+ - [Considerations for Using the Data](#considerations-for-using-the-data)
42
+ - [Social Impact of Dataset](#social-impact-of-dataset)
43
+ - [Discussion of Biases](#discussion-of-biases)
44
+ - [Other Known Limitations](#other-known-limitations)
45
+ - [Additional Information](#additional-information)
46
+ - [Dataset Curators](#dataset-curators)
47
+ - [Licensing Information](#licensing-information)
48
+ - [Citation Information](#citation-information)
49
+ - [Contributions](#contributions)
50
+
51
+ ## Dataset Description
52
+
53
+ - **Homepage:** [https://gov-report-data.github.io](https://gov-report-data.github.io)
54
+ - **Repository:** [https://github.com/luyang-huang96/LongDocSum](https://github.com/luyang-huang96/LongDocSum)
55
+ - **Paper:** [https://aclanthology.org/2021.naacl-main.112/](https://aclanthology.org/2021.naacl-main.112/)
56
+ - **Leaderboard:** [Needs More Information]
57
+ - **Point of Contact:** [Needs More Information]
58
+
59
+ ### Dataset Summary
60
+
61
+ Government report dataset consists of reports and associated summaries written by government research agencies including Congressional Research Service and U.S. Government Accountability Office.
62
+
63
+ Compared with other long document summarization datasets, government report dataset has longer summaries and documents and requires reading in more context to cover salient words to be summarized.
64
+
65
+ ### Supported Tasks and Leaderboards
66
+
67
+ [More Information Needed]
68
+
69
+ ### Languages
70
+
71
+ English
72
+
73
+ ## Dataset Structure
74
+
75
+ Three configs are available:
76
+ - **plain_text** (default): the text-to-text summarization setting used as in the original paper.
77
+ - **plain_text_with_recommendations**: the text-to-text summarization setting, with "What GAO recommends" included in the summary.
78
+ - **structure**: data with the section structure.
79
+
80
+ To use different configs, set the `name` argument of the `load_dataset` function.
81
+
82
+ ### Data Instances
83
+
84
+ #### plain_text & plain_text_with_recommendations
85
+
86
+ An example looks as follows.
87
+ ```
88
+ {
89
+ "id": "GAO_123456",
90
+ "document": "This is a test document.",
91
+ "summary": "This is a test summary"
92
+ }
93
+ ```
94
+
95
+ #### structure
96
+
97
+ An example looks as follows.
98
+ ```
99
+ {
100
+ "id": "GAO_123456",
101
+ "document_sections": {
102
+ "title": ["test docment section 1 title", "test docment section 1.1 title"],
103
+ "paragraphs": ["test document\nsection 1 paragraphs", "test document\nsection 1.1 paragraphs"],
104
+ "depth": [1, 2]
105
+ },
106
+ "summary_sections": {
107
+ "title": ["test summary section 1 title", "test summary section 2 title"],
108
+ "paragraphs": ["test summary\nsection 1 paragraphs", "test summary\nsection 2 paragraphs"]
109
+ }
110
+ }
111
+ ```
112
+
113
+ ### Data Fields
114
+
115
+ #### plain_text & plain_text_with_recommendations
116
+
117
+ - `id`: a `string` feature.
118
+ - `document`: a `string` feature.
119
+ - `summary`: a `string` feature.
120
+
121
+ #### structure
122
+
123
+ - `id`: a `string` feature.
124
+ - `document_sections`: a dictionary feature containing lists of (each element corresponds to a section):
125
+ - `title`: a `string` feature.
126
+ - `paragraphs`: a of `string` feature, with `\n` separating different paragraphs.
127
+ - `depth`: a `int32` feature.
128
+ - `summary_sections`: a dictionary feature containing lists of (each element corresponds to a section):
129
+ - `title`: a `string` feature.
130
+ - `paragraphs`: a `string` feature, with `\n` separating different paragraphs.
131
+
132
+ ### Data Splits
133
+
134
+ - train: 17519
135
+ - valid: 974
136
+ - test: 973
137
+
138
+ ## Dataset Creation
139
+
140
+ ### Curation Rationale
141
+
142
+ [More Information Needed]
143
+
144
+ ### Source Data
145
+
146
+ #### Initial Data Collection and Normalization
147
+
148
+ [More Information Needed]
149
+
150
+ #### Who are the source language producers?
151
+
152
+ Editors of the Congressional Research Service and U.S. Government Accountability Office.
153
+
154
+ ### Personal and Sensitive Information
155
+
156
+ None.
157
+
158
+ ## Considerations for Using the Data
159
+
160
+ ### Social Impact of Dataset
161
+
162
+ [More Information Needed]
163
+
164
+ ### Discussion of Biases
165
+
166
+ [More Information Needed]
167
+
168
+ ### Other Known Limitations
169
+
170
+ [More Information Needed]
171
+
172
+ ## Additional Information
173
+
174
+ ### Dataset Curators
175
+
176
+ [More Information Needed]
177
+
178
+ ### Licensing Information
179
+
180
+ CC BY 4.0
181
+
182
+ ### Citation Information
183
+
184
+ ```
185
+ @inproceedings{huang-etal-2021-efficient,
186
+ title = "Efficient Attentions for Long Document Summarization",
187
+ author = "Huang, Luyang and
188
+ Cao, Shuyang and
189
+ Parulian, Nikolaus and
190
+ Ji, Heng and
191
+ Wang, Lu",
192
+ booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
193
+ month = jun,
194
+ year = "2021",
195
+ address = "Online",
196
+ publisher = "Association for Computational Linguistics",
197
+ url = "https://aclanthology.org/2021.naacl-main.112",
198
+ doi = "10.18653/v1/2021.naacl-main.112",
199
+ pages = "1419--1436",
200
+ abstract = "The quadratic computational and memory complexities of large Transformers have limited their scalability for long document summarization. In this paper, we propose Hepos, a novel efficient encoder-decoder attention with head-wise positional strides to effectively pinpoint salient information from the source. We further conduct a systematic study of existing efficient self-attentions. Combined with Hepos, we are able to process ten times more tokens than existing models that use full attentions. For evaluation, we present a new dataset, GovReport, with significantly longer documents and summaries. Results show that our models produce significantly higher ROUGE scores than competitive comparisons, including new state-of-the-art results on PubMed. Human evaluation also shows that our models generate more informative summaries with fewer unfaithful errors.",
201
+ }
202
+ ```
data/crs_test.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bef184f1f4249ee4f145214b9f836a277152d5074395b879bd9ea891c519dece
3
+ size 19779267
data/crs_train.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5de865c921b76c4f1bc5f7f73fd38f9eeeef3ab7fe7cff205089853f079f74c6
3
+ size 359025392
data/crs_valid.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f56ff5bdbf381c053f106deadd566ad0ae3324b78c0d6ebe83c69121ecbdd24a
3
+ size 22496525
data/gao_test.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5285ee273ae430003c1e8fdb45ed7c6757349c114d2cae45160845ef60a49a94
3
+ size 39973256
data/gao_train.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee7a981a768c7f0c68d635e4a3cbfac03b95e45284181232d6c7021d511ea729
3
+ size 709026557
data/gao_valid.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cae7d5fecbcb507d1e84ecf8d476f26796d50a6690ac7c79fabe7b0e13a9b16
3
+ size 41604401
gov_report.py ADDED
@@ -0,0 +1,225 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """GovReport: The Government Report Long Document Summarization Dataset."""
2
+
3
+
4
+ import json
5
+
6
+ import datasets
7
+
8
+
9
+ logger = datasets.logging.get_logger(__name__)
10
+
11
+
12
+ _CITATION = """\
13
+ @inproceedings{huang-etal-2021-efficient,
14
+ title = "Efficient Attentions for Long Document Summarization",
15
+ author = "Huang, Luyang and
16
+ Cao, Shuyang and
17
+ Parulian, Nikolaus and
18
+ Ji, Heng and
19
+ Wang, Lu",
20
+ booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
21
+ month = jun,
22
+ year = "2021",
23
+ address = "Online",
24
+ publisher = "Association for Computational Linguistics",
25
+ url = "https://aclanthology.org/2021.naacl-main.112",
26
+ doi = "10.18653/v1/2021.naacl-main.112",
27
+ pages = "1419--1436",
28
+ abstract = "The quadratic computational and memory complexities of large Transformers have limited their scalability for long document summarization. In this paper, we propose Hepos, a novel efficient encoder-decoder attention with head-wise positional strides to effectively pinpoint salient information from the source. We further conduct a systematic study of existing efficient self-attentions. Combined with Hepos, we are able to process ten times more tokens than existing models that use full attentions. For evaluation, we present a new dataset, GovReport, with significantly longer documents and summaries. Results show that our models produce significantly higher ROUGE scores than competitive comparisons, including new state-of-the-art results on PubMed. Human evaluation also shows that our models generate more informative summaries with fewer unfaithful errors.",
29
+ }
30
+ """
31
+
32
+ _DESCRIPTION = """\
33
+ GovReport long document summarization dataset.
34
+
35
+ There are three configs:
36
+ - plain_text: plain text document-to-summary pairs
37
+ - plain_text_with_recommendations: plain text doucment-summary pairs, with "What GAO recommends" included in the summary
38
+ - structure: data with section structure
39
+ """
40
+
41
+ _URL = "https://huggingface.co/datasets/shuyangcao/gov_report/resolve/main/data/"
42
+ _URLS = {
43
+ "gao_train": _URL + "gao_train.jsonl",
44
+ "gao_valid": _URL + "gao_valid.jsonl",
45
+ "gao_test": _URL + "gao_test.jsonl",
46
+ "crs_train": _URL + "crs_train.jsonl",
47
+ "crs_valid": _URL + "crs_valid.jsonl",
48
+ "crs_test": _URL + "crs_test.jsonl",
49
+ }
50
+
51
+
52
+ def _recursive_load(section, keep_letter=False, depth=0):
53
+ sections = []
54
+ if section["section_title"] != "Letter" or (section["section_title"] == "Letter" and keep_letter):
55
+ sections.append({
56
+ "title": section["section_title"].strip(),
57
+ "paragraphs": "\n".join(section["paragraphs"]),
58
+ "depth": depth
59
+ })
60
+ for subsection in section["subsections"]:
61
+ child_sections = _recursive_load(subsection, keep_letter, depth + 1)
62
+ sections.extend(child_sections)
63
+ else:
64
+ for subsection in section["subsections"]:
65
+ child_sections = _recursive_load(subsection, keep_letter, depth)
66
+ sections.extend(child_sections)
67
+
68
+ return sections
69
+
70
+
71
+ class GovReportConfig(datasets.BuilderConfig):
72
+ """BuilderConfig for GovReport."""
73
+
74
+ def __init__(self, **kwargs):
75
+ """BuilderConfig for GovReport.
76
+ Args:
77
+ **kwargs: keyword arguments forwarded to super.
78
+ """
79
+ super(GovReportConfig, self).__init__(**kwargs)
80
+
81
+
82
+ class GovReport(datasets.GeneratorBasedBuilder):
83
+ VERSION = datasets.Version("1.0.0")
84
+
85
+ DEFAULT_CONFIG_NAME = "plain_text"
86
+
87
+ BUILDER_CONFIGS = [
88
+ GovReportConfig(
89
+ name="plain_text",
90
+ version=VERSION,
91
+ description="Plain text",
92
+ ),
93
+ GovReportConfig(
94
+ name="plain_text_with_recommendations",
95
+ version=VERSION,
96
+ description="Plain text with GAO recommendations",
97
+ ),
98
+ GovReportConfig(
99
+ name="structure",
100
+ version=VERSION,
101
+ description="structure data",
102
+ )
103
+ ]
104
+
105
+ def _info(self):
106
+ if self.config.name in ["plain_text", "plain_text_with_recommendations"]:
107
+ features = datasets.Features(
108
+ {
109
+ "id": datasets.Value("string"),
110
+ "document": datasets.Value("string"),
111
+ "summary": datasets.Value("string")
112
+ }
113
+ )
114
+ elif self.config.name == "structure":
115
+ features = datasets.Features(
116
+ {
117
+ "id": datasets.Value("string"),
118
+ "document_sections": datasets.features.Sequence(
119
+ {
120
+ "title": datasets.Value("string"),
121
+ "paragraphs": datasets.Value("string"),
122
+ "depth": datasets.Value("int32"),
123
+ }
124
+ ),
125
+ "summary_sections": datasets.features.Sequence(
126
+ {
127
+ "title": datasets.Value("string"),
128
+ "paragraphs": datasets.Value("string"),
129
+ }
130
+ ),
131
+ }
132
+ )
133
+ else:
134
+ raise ValueError("Unsupported config name {}".format(self.config.name))
135
+
136
+ return datasets.DatasetInfo(
137
+ description=_DESCRIPTION,
138
+ features=features,
139
+ supervised_keys=None,
140
+ homepage="",
141
+ citation=_CITATION,
142
+ )
143
+
144
+ def _split_generators(self, dl_manager):
145
+ downloaded_files = dl_manager.download_and_extract(_URLS)
146
+
147
+ return [
148
+ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"gao_filepath": downloaded_files["gao_train"], "crs_filepath": downloaded_files["crs_train"]}),
149
+ datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"gao_filepath": downloaded_files["gao_valid"], "crs_filepath": downloaded_files["crs_valid"]}),
150
+ datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"gao_filepath": downloaded_files["gao_test"], "crs_filepath": downloaded_files["crs_test"]}),
151
+ ]
152
+
153
+ def _generate_examples(self, gao_filepath, crs_filepath):
154
+ """This function returns the examples in the raw (text) form."""
155
+ logger.info(f"generating examples from = (GAO) {gao_filepath} and (CRS) {crs_filepath}")
156
+
157
+ with open(gao_filepath, "r") as f:
158
+ for line in f:
159
+ line = line.strip()
160
+ if not line:
161
+ continue
162
+ data = json.loads(line)
163
+
164
+ _id = 'GAO_' + data["id"]
165
+
166
+ document_sections = []
167
+ for lv1_section in data["report"]:
168
+ document_sections.extend(_recursive_load(lv1_section, keep_letter=False, depth=1))
169
+ summary_sections = [
170
+ {
171
+ "title": " ".join(highlight_section["section_title"].strip().split()),
172
+ "paragraphs": "\n".join([" ".join(paragraph.strip().split()) for paragraph in highlight_section["paragraphs"]])
173
+ } for highlight_section in data["highlight"]
174
+ ]
175
+
176
+ if self.config.name == "plain_text":
177
+ yield _id, {
178
+ "id": _id,
179
+ "document": " ".join([section["title"] + " " + section["paragraphs"] if section["paragraphs"] else section["title"] for section in document_sections]).replace("\n", " ").strip(),
180
+ "summary": " ".join([section["paragraphs"] for section in summary_sections if section["title"] != "What GAO Recommends"]).replace("\n", " ").strip(),
181
+ }
182
+ elif self.config.name == "plain_text_with_recommendations":
183
+ yield _id, {
184
+ "id": _id,
185
+ "document": " ".join([section["title"] + " " + section["paragraphs"] if section["paragraphs"] else section["title"] for section in document_sections]).replace("\n", " ").strip(),
186
+ "summary": " ".join([section["paragraphs"] for section in summary_sections]).replace("\n", " ").strip(),
187
+ }
188
+ elif self.config.name == "structure":
189
+ yield _id, {
190
+ "id": _id,
191
+ "document_sections": document_sections,
192
+ "summary_sections": summary_sections
193
+ }
194
+ else:
195
+ raise ValueError("Unsupported config name {}".format(self.config.name))
196
+
197
+ with open(crs_filepath, "r") as f:
198
+ for line in f:
199
+ line = line.strip()
200
+ if not line:
201
+ continue
202
+ data = json.loads(line)
203
+
204
+ _id = 'CRS_' + data["id"]
205
+
206
+ document_sections = _recursive_load(data["reports"], keep_letter=True, depth=0)
207
+ summary_sections = [{
208
+ "title": "",
209
+ "paragraphs": "\n".join([" ".join(paragraph.strip().split()) for paragraph in data["summary"]])
210
+ }]
211
+
212
+ if self.config.name in ["plain_text", "plain_text_with_recommendations"]:
213
+ yield _id, {
214
+ "id": _id,
215
+ "document": " ".join([section["title"] + " " + section["paragraphs"] if section["paragraphs"] else section["title"] for section in document_sections]).replace("\n", " ").strip(),
216
+ "summary": " ".join([section["paragraphs"] for section in summary_sections]).replace("\n", " ").strip(),
217
+ }
218
+ elif self.config.name == "structure":
219
+ yield _id, {
220
+ "id": _id,
221
+ "document_sections": document_sections,
222
+ "summary_sections": summary_sections
223
+ }
224
+ else:
225
+ raise ValueError("Unsupported config name {}".format(self.config.name))