Datasets:
Tasks:
Text Classification
Sub-tasks:
multi-class-classification
Languages:
English
Size:
10K - 100K
ArXiv:
Tags:
relation extraction
License:
Add data loading script and README.md
Browse files
README.md
ADDED
@@ -0,0 +1,334 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
annotations_creators:
|
3 |
+
- other
|
4 |
+
language:
|
5 |
+
- en
|
6 |
+
language_creators:
|
7 |
+
- found
|
8 |
+
license:
|
9 |
+
- other
|
10 |
+
multilinguality:
|
11 |
+
- monolingual
|
12 |
+
pretty_name: KBP37 is an English Relation Classification dataset
|
13 |
+
size_categories:
|
14 |
+
- 10K<n<100K
|
15 |
+
source_datasets:
|
16 |
+
- extended|other
|
17 |
+
tags:
|
18 |
+
- relation extraction
|
19 |
+
task_categories:
|
20 |
+
- text-classification
|
21 |
+
task_ids:
|
22 |
+
- multi-class-classification
|
23 |
+
dataset_info:
|
24 |
+
- config_name: kbp37
|
25 |
+
features:
|
26 |
+
- name: id
|
27 |
+
dtype: string
|
28 |
+
- name: sentence
|
29 |
+
dtype: string
|
30 |
+
- name: relation
|
31 |
+
dtype:
|
32 |
+
class_label:
|
33 |
+
names:
|
34 |
+
'0': no_relation
|
35 |
+
'1': org:alternate_names(e1,e2)
|
36 |
+
'2': org:alternate_names(e2,e1)
|
37 |
+
'3': org:city_of_headquarters(e1,e2)
|
38 |
+
'4': org:city_of_headquarters(e2,e1)
|
39 |
+
'5': org:country_of_headquarters(e1,e2)
|
40 |
+
'6': org:country_of_headquarters(e2,e1)
|
41 |
+
'7': org:founded(e1,e2)
|
42 |
+
'8': org:founded(e2,e1)
|
43 |
+
'9': org:founded_by(e1,e2)
|
44 |
+
'10': org:founded_by(e2,e1)
|
45 |
+
'11': org:members(e1,e2)
|
46 |
+
'12': org:members(e2,e1)
|
47 |
+
'13': org:stateorprovince_of_headquarters(e1,e2)
|
48 |
+
'14': org:stateorprovince_of_headquarters(e2,e1)
|
49 |
+
'15': org:subsidiaries(e1,e2)
|
50 |
+
'16': org:subsidiaries(e2,e1)
|
51 |
+
'17': org:top_members/employees(e1,e2)
|
52 |
+
'18': org:top_members/employees(e2,e1)
|
53 |
+
'19': per:alternate_names(e1,e2)
|
54 |
+
'20': per:alternate_names(e2,e1)
|
55 |
+
'21': per:cities_of_residence(e1,e2)
|
56 |
+
'22': per:cities_of_residence(e2,e1)
|
57 |
+
'23': per:countries_of_residence(e1,e2)
|
58 |
+
'24': per:countries_of_residence(e2,e1)
|
59 |
+
'25': per:country_of_birth(e1,e2)
|
60 |
+
'26': per:country_of_birth(e2,e1)
|
61 |
+
'27': per:employee_of(e1,e2)
|
62 |
+
'28': per:employee_of(e2,e1)
|
63 |
+
'29': per:origin(e1,e2)
|
64 |
+
'30': per:origin(e2,e1)
|
65 |
+
'31': per:spouse(e1,e2)
|
66 |
+
'32': per:spouse(e2,e1)
|
67 |
+
'33': per:stateorprovinces_of_residence(e1,e2)
|
68 |
+
'34': per:stateorprovinces_of_residence(e2,e1)
|
69 |
+
'35': per:title(e1,e2)
|
70 |
+
'36': per:title(e2,e1)
|
71 |
+
splits:
|
72 |
+
- name: train
|
73 |
+
num_bytes: 3570626
|
74 |
+
num_examples: 15917
|
75 |
+
- name: validation
|
76 |
+
num_bytes: 388935
|
77 |
+
num_examples: 1724
|
78 |
+
- name: test
|
79 |
+
num_bytes: 762806
|
80 |
+
num_examples: 3405
|
81 |
+
download_size: 5106673
|
82 |
+
dataset_size: 4722367
|
83 |
+
- config_name: kbp37_formatted
|
84 |
+
features:
|
85 |
+
- name: id
|
86 |
+
dtype: string
|
87 |
+
- name: token
|
88 |
+
sequence: string
|
89 |
+
- name: subj_start
|
90 |
+
dtype: int32
|
91 |
+
- name: subj_end
|
92 |
+
dtype: int32
|
93 |
+
- name: obj_start
|
94 |
+
dtype: int32
|
95 |
+
- name: obj_end
|
96 |
+
dtype: int32
|
97 |
+
- name: relation
|
98 |
+
dtype:
|
99 |
+
class_label:
|
100 |
+
names:
|
101 |
+
'0': no_relation
|
102 |
+
'1': org:alternate_names(e1,e2)
|
103 |
+
'2': org:alternate_names(e2,e1)
|
104 |
+
'3': org:city_of_headquarters(e1,e2)
|
105 |
+
'4': org:city_of_headquarters(e2,e1)
|
106 |
+
'5': org:country_of_headquarters(e1,e2)
|
107 |
+
'6': org:country_of_headquarters(e2,e1)
|
108 |
+
'7': org:founded(e1,e2)
|
109 |
+
'8': org:founded(e2,e1)
|
110 |
+
'9': org:founded_by(e1,e2)
|
111 |
+
'10': org:founded_by(e2,e1)
|
112 |
+
'11': org:members(e1,e2)
|
113 |
+
'12': org:members(e2,e1)
|
114 |
+
'13': org:stateorprovince_of_headquarters(e1,e2)
|
115 |
+
'14': org:stateorprovince_of_headquarters(e2,e1)
|
116 |
+
'15': org:subsidiaries(e1,e2)
|
117 |
+
'16': org:subsidiaries(e2,e1)
|
118 |
+
'17': org:top_members/employees(e1,e2)
|
119 |
+
'18': org:top_members/employees(e2,e1)
|
120 |
+
'19': per:alternate_names(e1,e2)
|
121 |
+
'20': per:alternate_names(e2,e1)
|
122 |
+
'21': per:cities_of_residence(e1,e2)
|
123 |
+
'22': per:cities_of_residence(e2,e1)
|
124 |
+
'23': per:countries_of_residence(e1,e2)
|
125 |
+
'24': per:countries_of_residence(e2,e1)
|
126 |
+
'25': per:country_of_birth(e1,e2)
|
127 |
+
'26': per:country_of_birth(e2,e1)
|
128 |
+
'27': per:employee_of(e1,e2)
|
129 |
+
'28': per:employee_of(e2,e1)
|
130 |
+
'29': per:origin(e1,e2)
|
131 |
+
'30': per:origin(e2,e1)
|
132 |
+
'31': per:spouse(e1,e2)
|
133 |
+
'32': per:spouse(e2,e1)
|
134 |
+
'33': per:stateorprovinces_of_residence(e1,e2)
|
135 |
+
'34': per:stateorprovinces_of_residence(e2,e1)
|
136 |
+
'35': per:title(e1,e2)
|
137 |
+
'36': per:title(e2,e1)
|
138 |
+
splits:
|
139 |
+
- name: train
|
140 |
+
num_bytes: 4975792
|
141 |
+
num_examples: 15917
|
142 |
+
- name: validation
|
143 |
+
num_bytes: 542576
|
144 |
+
num_examples: 1724
|
145 |
+
- name: test
|
146 |
+
num_bytes: 1062977
|
147 |
+
num_examples: 3405
|
148 |
+
download_size: 5106673
|
149 |
+
dataset_size: 6581345
|
150 |
+
---
|
151 |
+
# Dataset Card for "kbp37"
|
152 |
+
## Table of Contents
|
153 |
+
- [Table of Contents](#table-of-contents)
|
154 |
+
- [Dataset Description](#dataset-description)
|
155 |
+
- [Dataset Summary](#dataset-summary)
|
156 |
+
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
|
157 |
+
- [Languages](#languages)
|
158 |
+
- [Dataset Structure](#dataset-structure)
|
159 |
+
- [Data Instances](#data-instances)
|
160 |
+
- [Data Fields](#data-fields)
|
161 |
+
- [Data Splits](#data-splits)
|
162 |
+
- [Dataset Creation](#dataset-creation)
|
163 |
+
- [Curation Rationale](#curation-rationale)
|
164 |
+
- [Source Data](#source-data)
|
165 |
+
- [Annotations](#annotations)
|
166 |
+
- [Personal and Sensitive Information](#personal-and-sensitive-information)
|
167 |
+
- [Considerations for Using the Data](#considerations-for-using-the-data)
|
168 |
+
- [Social Impact of Dataset](#social-impact-of-dataset)
|
169 |
+
- [Discussion of Biases](#discussion-of-biases)
|
170 |
+
- [Other Known Limitations](#other-known-limitations)
|
171 |
+
- [Additional Information](#additional-information)
|
172 |
+
- [Dataset Curators](#dataset-curators)
|
173 |
+
- [Licensing Information](#licensing-information)
|
174 |
+
- [Citation Information](#citation-information)
|
175 |
+
- [Contributions](#contributions)
|
176 |
+
|
177 |
+
## Dataset Description
|
178 |
+
- **Homepage:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
179 |
+
- **Repository:** [kbp37](https://github.com/zhangdongxu/kbp37)
|
180 |
+
- **Paper:** [Relation Classification via Recurrent Neural Network](https://arxiv.org/abs/1508.01006)
|
181 |
+
- **Point of Contact:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
182 |
+
- **Size of downloaded dataset files:** 5.11 MB
|
183 |
+
- **Size of the generated dataset:** 6.58 MB
|
184 |
+
|
185 |
+
### Dataset Summary
|
186 |
+
KBP37 is a revision of MIML-RE annotation dataset, provided by Gabor Angeli et al. (2014). They use both the 2010 and
|
187 |
+
2013 KBP official document collections, as well as a July 2013 dump of Wikipedia as the text corpus for annotation.
|
188 |
+
There are 33811 sentences been annotated. Zhang and Wang made several refinements:
|
189 |
+
1. They add direction to the relation names, e.g. '`per:employee_of`' is split into '`per:employee of(e1,e2)`'
|
190 |
+
and '`per:employee of(e2,e1)`'. They also replace '`org:parents`' with '`org:subsidiaries`' and replace
|
191 |
+
'`org:member of’ with '`org:member`' (by their reverse directions).
|
192 |
+
2. They discard low frequency relations such that both directions of each relation occur more than 100 times in the
|
193 |
+
dataset.
|
194 |
+
|
195 |
+
KBP37 contains 18 directional relations and an additional '`no_relation`' relation, resulting in 37 relation classes.
|
196 |
+
|
197 |
+
Note:
|
198 |
+
- There is a formatted version that you can load with `datasets.load_dataset('kbp37', name='kbp37_formatted')`. This version is tokenized with spaCy and provides entity offsets.
|
199 |
+
|
200 |
+
|
201 |
+
### Supported Tasks and Leaderboards
|
202 |
+
|
203 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
204 |
+
|
205 |
+
### Languages
|
206 |
+
|
207 |
+
The language data in KBP37 is in English (BCP-47 en)
|
208 |
+
|
209 |
+
## Dataset Structure
|
210 |
+
|
211 |
+
### Data Instances
|
212 |
+
|
213 |
+
#### kbp37
|
214 |
+
An example of 'train' looks as follows:
|
215 |
+
```json
|
216 |
+
{
|
217 |
+
"id": "0",
|
218 |
+
"sentence": "<e1> Thom Yorke </e1> of <e2> Radiohead </e2> has included the + for many of his signature distortion sounds using a variety of guitars to achieve various tonal options .",
|
219 |
+
"relation": 27
|
220 |
+
}
|
221 |
+
```
|
222 |
+
|
223 |
+
#### kbp37_formatted
|
224 |
+
An example of 'train' looks as follows:
|
225 |
+
```json
|
226 |
+
{
|
227 |
+
"id": "1",
|
228 |
+
"token": ["Leland", "High", "School", "is", "a", "public", "high", "school", "located", "in", "the", "Almaden", "Valley", "in", "San", "Jose", "California", "USA", "in", "the", "San", "Jose", "Unified", "School", "District", "."],
|
229 |
+
"subj_start": 0,
|
230 |
+
"subj_end": 3,
|
231 |
+
"obj_start": 14,
|
232 |
+
"obj_end": 16,
|
233 |
+
"relation": 3
|
234 |
+
}
|
235 |
+
```
|
236 |
+
|
237 |
+
### Data Fields
|
238 |
+
|
239 |
+
#### kbp37
|
240 |
+
- `id`: the instance id of this sentence, a `string` feature.
|
241 |
+
- `sentence`: the sentence, a `string` features.
|
242 |
+
- `relation`: the relation label of this instance, a `string` classification label.
|
243 |
+
|
244 |
+
#### kbp37_formatted
|
245 |
+
- `id`: the instance id of this sentence, a `string` feature.
|
246 |
+
- `token`: the list of tokens of this sentence, obtained with spaCy, a `list` of `string` features.
|
247 |
+
- `subj_start`: the 0-based index of the start token of the relation subject mention, an `ìnt` feature.
|
248 |
+
- `subj_end`: the 0-based index of the end token of the relation subject mention, exclusive, an `ìnt` feature.
|
249 |
+
- `obj_start`: the 0-based index of the start token of the relation object mention, an `ìnt` feature.
|
250 |
+
- `obj_end`: the 0-based index of the end token of the relation object mention, exclusive, an `ìnt` feature.
|
251 |
+
- `relation`: the relation label of this instance, a `string` classification label.
|
252 |
+
|
253 |
+
### Data Splits
|
254 |
+
|
255 |
+
| | Train | Dev | Test |
|
256 |
+
|-------|-------|------|------|
|
257 |
+
| KBP37 | 15917 | 1724 | 3405 |
|
258 |
+
|
259 |
+
## Dataset Creation
|
260 |
+
|
261 |
+
### Curation Rationale
|
262 |
+
|
263 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
264 |
+
|
265 |
+
### Source Data
|
266 |
+
|
267 |
+
#### Initial Data Collection and Normalization
|
268 |
+
|
269 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
270 |
+
|
271 |
+
#### Who are the source language producers?
|
272 |
+
|
273 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
274 |
+
|
275 |
+
### Annotations
|
276 |
+
|
277 |
+
#### Annotation process
|
278 |
+
|
279 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
280 |
+
|
281 |
+
#### Who are the annotators?
|
282 |
+
|
283 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
284 |
+
|
285 |
+
### Personal and Sensitive Information
|
286 |
+
|
287 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
288 |
+
|
289 |
+
## Considerations for Using the Data
|
290 |
+
|
291 |
+
### Social Impact of Dataset
|
292 |
+
|
293 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
294 |
+
|
295 |
+
### Discussion of Biases
|
296 |
+
|
297 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
298 |
+
|
299 |
+
### Other Known Limitations
|
300 |
+
|
301 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
302 |
+
|
303 |
+
## Additional Information
|
304 |
+
|
305 |
+
### Dataset Curators
|
306 |
+
|
307 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
308 |
+
|
309 |
+
### Licensing Information
|
310 |
+
|
311 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
312 |
+
|
313 |
+
### Citation Information
|
314 |
+
|
315 |
+
```
|
316 |
+
@article{DBLP:journals/corr/ZhangW15a,
|
317 |
+
author = {Dongxu Zhang and
|
318 |
+
Dong Wang},
|
319 |
+
title = {Relation Classification via Recurrent Neural Network},
|
320 |
+
journal = {CoRR},
|
321 |
+
volume = {abs/1508.01006},
|
322 |
+
year = {2015},
|
323 |
+
url = {http://arxiv.org/abs/1508.01006},
|
324 |
+
eprinttype = {arXiv},
|
325 |
+
eprint = {1508.01006},
|
326 |
+
timestamp = {Fri, 04 Nov 2022 18:37:50 +0100},
|
327 |
+
biburl = {https://dblp.org/rec/journals/corr/ZhangW15a.bib},
|
328 |
+
bibsource = {dblp computer science bibliography, https://dblp.org}
|
329 |
+
}
|
330 |
+
```
|
331 |
+
|
332 |
+
### Contributions
|
333 |
+
|
334 |
+
Thanks to [@phucdev](https://github.com/phucdev) for adding this dataset.
|
kbp37.py
ADDED
@@ -0,0 +1,234 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# coding=utf-8
|
2 |
+
# Copyright 2022 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 |
+
"""The KBP37 dataset for English Relation Classification"""
|
17 |
+
|
18 |
+
import datasets
|
19 |
+
|
20 |
+
_CITATION = """\
|
21 |
+
@article{DBLP:journals/corr/ZhangW15a,
|
22 |
+
author = {Dongxu Zhang and
|
23 |
+
Dong Wang},
|
24 |
+
title = {Relation Classification via Recurrent Neural Network},
|
25 |
+
journal = {CoRR},
|
26 |
+
volume = {abs/1508.01006},
|
27 |
+
year = {2015},
|
28 |
+
url = {http://arxiv.org/abs/1508.01006},
|
29 |
+
eprinttype = {arXiv},
|
30 |
+
eprint = {1508.01006},
|
31 |
+
timestamp = {Fri, 04 Nov 2022 18:37:50 +0100},
|
32 |
+
biburl = {https://dblp.org/rec/journals/corr/ZhangW15a.bib},
|
33 |
+
bibsource = {dblp computer science bibliography, https://dblp.org}
|
34 |
+
}
|
35 |
+
"""
|
36 |
+
|
37 |
+
_DESCRIPTION = """\
|
38 |
+
KBP37 is a revision of MIML-RE annotation dataset, provided by Gabor Angeli et al. (2014). They use both the 2010 and
|
39 |
+
2013 KBP official document collections, as well as a July 2013 dump of Wikipedia as the text corpus for annotation.
|
40 |
+
There are 33811 sentences been annotated. Zhang and Wang made several refinements:
|
41 |
+
1. They add direction to the relation names, e.g. '`per:employee_of`' is split into '`per:employee of(e1,e2)`'
|
42 |
+
and '`per:employee of(e2,e1)`'. They also replace '`org:parents`' with '`org:subsidiaries`' and replace
|
43 |
+
'`org:member of’ with '`org:member`' (by their reverse directions).
|
44 |
+
2. They discard low frequency relations such that both directions of each relation occur more than 100 times in the
|
45 |
+
dataset.
|
46 |
+
|
47 |
+
KBP37 contains 18 directional relations and an additional '`no_relation`' relation, resulting in 37 relation classes.
|
48 |
+
"""
|
49 |
+
|
50 |
+
_HOMEPAGE = ""
|
51 |
+
|
52 |
+
_LICENSE = ""
|
53 |
+
|
54 |
+
|
55 |
+
# The HuggingFace dataset library don't host the datasets but only point to the original files
|
56 |
+
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
57 |
+
_URLs = {
|
58 |
+
"train": "https://raw.githubusercontent.com/zhangdongxu/kbp37/master/train.txt",
|
59 |
+
"validation": "https://raw.githubusercontent.com/zhangdongxu/kbp37/master/dev.txt",
|
60 |
+
"test": "https://raw.githubusercontent.com/zhangdongxu/kbp37/master/test.txt"
|
61 |
+
}
|
62 |
+
|
63 |
+
_VERSION = datasets.Version("1.0.0")
|
64 |
+
|
65 |
+
_CLASS_LABELS = [
|
66 |
+
"no_relation",
|
67 |
+
"org:alternate_names(e1,e2)",
|
68 |
+
"org:alternate_names(e2,e1)",
|
69 |
+
"org:city_of_headquarters(e1,e2)",
|
70 |
+
"org:city_of_headquarters(e2,e1)",
|
71 |
+
"org:country_of_headquarters(e1,e2)",
|
72 |
+
"org:country_of_headquarters(e2,e1)",
|
73 |
+
"org:founded(e1,e2)",
|
74 |
+
"org:founded(e2,e1)",
|
75 |
+
"org:founded_by(e1,e2)",
|
76 |
+
"org:founded_by(e2,e1)",
|
77 |
+
"org:members(e1,e2)",
|
78 |
+
"org:members(e2,e1)",
|
79 |
+
"org:stateorprovince_of_headquarters(e1,e2)",
|
80 |
+
"org:stateorprovince_of_headquarters(e2,e1)",
|
81 |
+
"org:subsidiaries(e1,e2)",
|
82 |
+
"org:subsidiaries(e2,e1)",
|
83 |
+
"org:top_members/employees(e1,e2)",
|
84 |
+
"org:top_members/employees(e2,e1)",
|
85 |
+
"per:alternate_names(e1,e2)",
|
86 |
+
"per:alternate_names(e2,e1)",
|
87 |
+
"per:cities_of_residence(e1,e2)",
|
88 |
+
"per:cities_of_residence(e2,e1)",
|
89 |
+
"per:countries_of_residence(e1,e2)",
|
90 |
+
"per:countries_of_residence(e2,e1)",
|
91 |
+
"per:country_of_birth(e1,e2)",
|
92 |
+
"per:country_of_birth(e2,e1)",
|
93 |
+
"per:employee_of(e1,e2)",
|
94 |
+
"per:employee_of(e2,e1)",
|
95 |
+
"per:origin(e1,e2)",
|
96 |
+
"per:origin(e2,e1)",
|
97 |
+
"per:spouse(e1,e2)",
|
98 |
+
"per:spouse(e2,e1)",
|
99 |
+
"per:stateorprovinces_of_residence(e1,e2)",
|
100 |
+
"per:stateorprovinces_of_residence(e2,e1)",
|
101 |
+
"per:title(e1,e2)",
|
102 |
+
"per:title(e2,e1)"
|
103 |
+
]
|
104 |
+
|
105 |
+
|
106 |
+
class KBP37(datasets.GeneratorBasedBuilder):
|
107 |
+
"""KBP37 is a relation extraction dataset"""
|
108 |
+
|
109 |
+
BUILDER_CONFIGS = [
|
110 |
+
datasets.BuilderConfig(
|
111 |
+
name="kbp37", version=_VERSION, description="The KBP37 dataset."
|
112 |
+
),
|
113 |
+
datasets.BuilderConfig(
|
114 |
+
name="kbp37_formatted", version=_VERSION, description="The formatted KBP37 dataset."
|
115 |
+
)
|
116 |
+
]
|
117 |
+
|
118 |
+
DEFAULT_CONFIG_NAME = "kbp37" # type: ignore
|
119 |
+
|
120 |
+
def _info(self):
|
121 |
+
if self.config.name == "kbp37_formatted":
|
122 |
+
features = datasets.Features(
|
123 |
+
{
|
124 |
+
"id": datasets.Value("string"),
|
125 |
+
"token": datasets.Sequence(datasets.Value("string")),
|
126 |
+
"subj_start": datasets.Value("int32"),
|
127 |
+
"subj_end": datasets.Value("int32"),
|
128 |
+
"obj_start": datasets.Value("int32"),
|
129 |
+
"obj_end": datasets.Value("int32"),
|
130 |
+
"relation": datasets.ClassLabel(names=_CLASS_LABELS),
|
131 |
+
}
|
132 |
+
)
|
133 |
+
else:
|
134 |
+
features = datasets.Features(
|
135 |
+
{
|
136 |
+
"id": datasets.Value("string"),
|
137 |
+
"sentence": datasets.Value("string"),
|
138 |
+
"relation": datasets.ClassLabel(names=_CLASS_LABELS),
|
139 |
+
}
|
140 |
+
)
|
141 |
+
|
142 |
+
return datasets.DatasetInfo(
|
143 |
+
# This is the description that will appear on the datasets page.
|
144 |
+
description=_DESCRIPTION,
|
145 |
+
# This defines the different columns of the dataset and their types
|
146 |
+
features=features, # Here we define them above because they are different between the two configurations
|
147 |
+
# If there's a common (input, target) tuple from the features,
|
148 |
+
# specify them here. They'll be used if as_supervised=True in
|
149 |
+
# builder.as_dataset.
|
150 |
+
supervised_keys=None,
|
151 |
+
# Homepage of the dataset for documentation
|
152 |
+
homepage=_HOMEPAGE,
|
153 |
+
# License for the dataset if available
|
154 |
+
license=_LICENSE,
|
155 |
+
# Citation for the dataset
|
156 |
+
citation=_CITATION,
|
157 |
+
)
|
158 |
+
|
159 |
+
def _split_generators(self, dl_manager):
|
160 |
+
"""Returns SplitGenerators."""
|
161 |
+
# If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
|
162 |
+
|
163 |
+
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLs
|
164 |
+
# It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
|
165 |
+
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
|
166 |
+
|
167 |
+
downloaded_files = dl_manager.download_and_extract(_URLs)
|
168 |
+
|
169 |
+
return [datasets.SplitGenerator(name=i, gen_kwargs={"filepath": downloaded_files[str(i)]})
|
170 |
+
for i in [datasets.Split.TRAIN, datasets.Split.VALIDATION, datasets.Split.TEST]]
|
171 |
+
|
172 |
+
def _generate_examples(self, filepath):
|
173 |
+
"""Yields examples."""
|
174 |
+
# This method will receive as arguments the `gen_kwargs` defined in the previous `_split_generators` method.
|
175 |
+
# It is in charge of opening the given file and yielding (key, example) tuples from the dataset
|
176 |
+
# The key is not important, it's more here for legacy reason (legacy from tfds)
|
177 |
+
|
178 |
+
with open(filepath, encoding="utf-8") as f:
|
179 |
+
data = []
|
180 |
+
example_line = None
|
181 |
+
for idx, line in enumerate(f.readlines()):
|
182 |
+
line_no = idx % 4 # first line contains example, second line relation, third and fourth lines are \n
|
183 |
+
if line_no == 0:
|
184 |
+
example_line = line.strip().split("\t")
|
185 |
+
elif line_no == 1:
|
186 |
+
data.append({"example": example_line, "relation": line.strip()})
|
187 |
+
for example in data:
|
188 |
+
id_ = example["example"][0]
|
189 |
+
text = example["example"][1]
|
190 |
+
assert text[:2] == "\" " and text[-2:] == " \""
|
191 |
+
text = text[2:-2]
|
192 |
+
relation = example["relation"]
|
193 |
+
|
194 |
+
if self.config.name == "kbp37_formatted":
|
195 |
+
text = text.replace("<e1>", " <e1> ")
|
196 |
+
text = text.replace("<e2>", " <e2> ")
|
197 |
+
text = text.replace("</e1>", " </e1> ")
|
198 |
+
text = text.replace("</e2>", " </e2> ")
|
199 |
+
text = text.strip().replace(r"\s\s+", r"\s")
|
200 |
+
tokens = text.split()
|
201 |
+
subj_start = tokens.index("<e1>")
|
202 |
+
obj_start = tokens.index("<e2>")
|
203 |
+
if subj_start < obj_start:
|
204 |
+
tokens.pop(subj_start)
|
205 |
+
subj_end = tokens.index("</e1>")
|
206 |
+
tokens.pop(subj_end)
|
207 |
+
obj_start = tokens.index("<e2>")
|
208 |
+
tokens.pop(obj_start)
|
209 |
+
obj_end = tokens.index("</e2>")
|
210 |
+
tokens.pop(obj_end)
|
211 |
+
else:
|
212 |
+
tokens.pop(obj_start)
|
213 |
+
obj_end = tokens.index("</e2>")
|
214 |
+
tokens.pop(obj_end)
|
215 |
+
subj_start = tokens.index("<e1>")
|
216 |
+
tokens.pop(subj_start)
|
217 |
+
subj_end = tokens.index("</e1>")
|
218 |
+
tokens.pop(subj_end)
|
219 |
+
|
220 |
+
yield int(id_), {
|
221 |
+
"id": id_,
|
222 |
+
"token": tokens,
|
223 |
+
"subj_start": subj_start,
|
224 |
+
"subj_end": subj_end,
|
225 |
+
"obj_start": obj_start,
|
226 |
+
"obj_end": obj_end,
|
227 |
+
"relation": relation,
|
228 |
+
}
|
229 |
+
else:
|
230 |
+
yield int(id_), {
|
231 |
+
"id": id_,
|
232 |
+
"sentence": text,
|
233 |
+
"relation": relation,
|
234 |
+
}
|