parquet-converter
commited on
Commit
•
b46a43b
1
Parent(s):
89a6001
Update parquet files
Browse files- README.md +0 -299
- dataset_infos.json +0 -1
- masked_medhop/qangaroo-train.parquet +3 -0
- masked_medhop/qangaroo-validation.parquet +3 -0
- masked_wikihop/qangaroo-train.parquet +3 -0
- masked_wikihop/qangaroo-validation.parquet +3 -0
- medhop/qangaroo-train.parquet +3 -0
- medhop/qangaroo-validation.parquet +3 -0
- qangaroo.py +0 -126
- wikihop/qangaroo-train.parquet +3 -0
- wikihop/qangaroo-validation.parquet +3 -0
README.md
DELETED
@@ -1,299 +0,0 @@
|
|
1 |
-
---
|
2 |
-
language:
|
3 |
-
- en
|
4 |
-
paperswithcode_id: null
|
5 |
-
pretty_name: qangaroo
|
6 |
-
dataset_info:
|
7 |
-
- config_name: medhop
|
8 |
-
features:
|
9 |
-
- name: query
|
10 |
-
dtype: string
|
11 |
-
- name: supports
|
12 |
-
sequence: string
|
13 |
-
- name: candidates
|
14 |
-
sequence: string
|
15 |
-
- name: answer
|
16 |
-
dtype: string
|
17 |
-
- name: id
|
18 |
-
dtype: string
|
19 |
-
splits:
|
20 |
-
- name: train
|
21 |
-
num_bytes: 93947725
|
22 |
-
num_examples: 1620
|
23 |
-
- name: validation
|
24 |
-
num_bytes: 16463555
|
25 |
-
num_examples: 342
|
26 |
-
download_size: 339843061
|
27 |
-
dataset_size: 110411280
|
28 |
-
- config_name: masked_medhop
|
29 |
-
features:
|
30 |
-
- name: query
|
31 |
-
dtype: string
|
32 |
-
- name: supports
|
33 |
-
sequence: string
|
34 |
-
- name: candidates
|
35 |
-
sequence: string
|
36 |
-
- name: answer
|
37 |
-
dtype: string
|
38 |
-
- name: id
|
39 |
-
dtype: string
|
40 |
-
splits:
|
41 |
-
- name: train
|
42 |
-
num_bytes: 95823986
|
43 |
-
num_examples: 1620
|
44 |
-
- name: validation
|
45 |
-
num_bytes: 16802484
|
46 |
-
num_examples: 342
|
47 |
-
download_size: 339843061
|
48 |
-
dataset_size: 112626470
|
49 |
-
- config_name: wikihop
|
50 |
-
features:
|
51 |
-
- name: query
|
52 |
-
dtype: string
|
53 |
-
- name: supports
|
54 |
-
sequence: string
|
55 |
-
- name: candidates
|
56 |
-
sequence: string
|
57 |
-
- name: answer
|
58 |
-
dtype: string
|
59 |
-
- name: id
|
60 |
-
dtype: string
|
61 |
-
splits:
|
62 |
-
- name: train
|
63 |
-
num_bytes: 325994029
|
64 |
-
num_examples: 43738
|
65 |
-
- name: validation
|
66 |
-
num_bytes: 40869634
|
67 |
-
num_examples: 5129
|
68 |
-
download_size: 339843061
|
69 |
-
dataset_size: 366863663
|
70 |
-
- config_name: masked_wikihop
|
71 |
-
features:
|
72 |
-
- name: query
|
73 |
-
dtype: string
|
74 |
-
- name: supports
|
75 |
-
sequence: string
|
76 |
-
- name: candidates
|
77 |
-
sequence: string
|
78 |
-
- name: answer
|
79 |
-
dtype: string
|
80 |
-
- name: id
|
81 |
-
dtype: string
|
82 |
-
splits:
|
83 |
-
- name: train
|
84 |
-
num_bytes: 348290479
|
85 |
-
num_examples: 43738
|
86 |
-
- name: validation
|
87 |
-
num_bytes: 43689810
|
88 |
-
num_examples: 5129
|
89 |
-
download_size: 339843061
|
90 |
-
dataset_size: 391980289
|
91 |
-
---
|
92 |
-
|
93 |
-
# Dataset Card for "qangaroo"
|
94 |
-
|
95 |
-
## Table of Contents
|
96 |
-
- [Dataset Description](#dataset-description)
|
97 |
-
- [Dataset Summary](#dataset-summary)
|
98 |
-
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
|
99 |
-
- [Languages](#languages)
|
100 |
-
- [Dataset Structure](#dataset-structure)
|
101 |
-
- [Data Instances](#data-instances)
|
102 |
-
- [Data Fields](#data-fields)
|
103 |
-
- [Data Splits](#data-splits)
|
104 |
-
- [Dataset Creation](#dataset-creation)
|
105 |
-
- [Curation Rationale](#curation-rationale)
|
106 |
-
- [Source Data](#source-data)
|
107 |
-
- [Annotations](#annotations)
|
108 |
-
- [Personal and Sensitive Information](#personal-and-sensitive-information)
|
109 |
-
- [Considerations for Using the Data](#considerations-for-using-the-data)
|
110 |
-
- [Social Impact of Dataset](#social-impact-of-dataset)
|
111 |
-
- [Discussion of Biases](#discussion-of-biases)
|
112 |
-
- [Other Known Limitations](#other-known-limitations)
|
113 |
-
- [Additional Information](#additional-information)
|
114 |
-
- [Dataset Curators](#dataset-curators)
|
115 |
-
- [Licensing Information](#licensing-information)
|
116 |
-
- [Citation Information](#citation-information)
|
117 |
-
- [Contributions](#contributions)
|
118 |
-
|
119 |
-
## Dataset Description
|
120 |
-
|
121 |
-
- **Homepage:** [http://qangaroo.cs.ucl.ac.uk/index.html](http://qangaroo.cs.ucl.ac.uk/index.html)
|
122 |
-
- **Repository:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
123 |
-
- **Paper:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
124 |
-
- **Point of Contact:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
125 |
-
- **Size of downloaded dataset files:** 1296.40 MB
|
126 |
-
- **Size of the generated dataset:** 936.40 MB
|
127 |
-
- **Total amount of disk used:** 2232.79 MB
|
128 |
-
|
129 |
-
### Dataset Summary
|
130 |
-
|
131 |
-
We have created two new Reading Comprehension datasets focussing on multi-hop (alias multi-step) inference.
|
132 |
-
|
133 |
-
Several pieces of information often jointly imply another fact. In multi-hop inference, a new fact is derived by combining facts via a chain of multiple steps.
|
134 |
-
|
135 |
-
Our aim is to build Reading Comprehension methods that perform multi-hop inference on text, where individual facts are spread out across different documents.
|
136 |
-
|
137 |
-
The two QAngaroo datasets provide a training and evaluation resource for such methods.
|
138 |
-
|
139 |
-
### Supported Tasks and Leaderboards
|
140 |
-
|
141 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
142 |
-
|
143 |
-
### Languages
|
144 |
-
|
145 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
146 |
-
|
147 |
-
## Dataset Structure
|
148 |
-
|
149 |
-
### Data Instances
|
150 |
-
|
151 |
-
#### masked_medhop
|
152 |
-
|
153 |
-
- **Size of downloaded dataset files:** 324.10 MB
|
154 |
-
- **Size of the generated dataset:** 107.41 MB
|
155 |
-
- **Total amount of disk used:** 431.51 MB
|
156 |
-
|
157 |
-
An example of 'validation' looks as follows.
|
158 |
-
```
|
159 |
-
|
160 |
-
```
|
161 |
-
|
162 |
-
#### masked_wikihop
|
163 |
-
|
164 |
-
- **Size of downloaded dataset files:** 324.10 MB
|
165 |
-
- **Size of the generated dataset:** 373.82 MB
|
166 |
-
- **Total amount of disk used:** 697.92 MB
|
167 |
-
|
168 |
-
An example of 'validation' looks as follows.
|
169 |
-
```
|
170 |
-
|
171 |
-
```
|
172 |
-
|
173 |
-
#### medhop
|
174 |
-
|
175 |
-
- **Size of downloaded dataset files:** 324.10 MB
|
176 |
-
- **Size of the generated dataset:** 105.30 MB
|
177 |
-
- **Total amount of disk used:** 429.40 MB
|
178 |
-
|
179 |
-
An example of 'validation' looks as follows.
|
180 |
-
```
|
181 |
-
|
182 |
-
```
|
183 |
-
|
184 |
-
#### wikihop
|
185 |
-
|
186 |
-
- **Size of downloaded dataset files:** 324.10 MB
|
187 |
-
- **Size of the generated dataset:** 349.87 MB
|
188 |
-
- **Total amount of disk used:** 673.97 MB
|
189 |
-
|
190 |
-
An example of 'validation' looks as follows.
|
191 |
-
```
|
192 |
-
|
193 |
-
```
|
194 |
-
|
195 |
-
### Data Fields
|
196 |
-
|
197 |
-
The data fields are the same among all splits.
|
198 |
-
|
199 |
-
#### masked_medhop
|
200 |
-
- `query`: a `string` feature.
|
201 |
-
- `supports`: a `list` of `string` features.
|
202 |
-
- `candidates`: a `list` of `string` features.
|
203 |
-
- `answer`: a `string` feature.
|
204 |
-
- `id`: a `string` feature.
|
205 |
-
|
206 |
-
#### masked_wikihop
|
207 |
-
- `query`: a `string` feature.
|
208 |
-
- `supports`: a `list` of `string` features.
|
209 |
-
- `candidates`: a `list` of `string` features.
|
210 |
-
- `answer`: a `string` feature.
|
211 |
-
- `id`: a `string` feature.
|
212 |
-
|
213 |
-
#### medhop
|
214 |
-
- `query`: a `string` feature.
|
215 |
-
- `supports`: a `list` of `string` features.
|
216 |
-
- `candidates`: a `list` of `string` features.
|
217 |
-
- `answer`: a `string` feature.
|
218 |
-
- `id`: a `string` feature.
|
219 |
-
|
220 |
-
#### wikihop
|
221 |
-
- `query`: a `string` feature.
|
222 |
-
- `supports`: a `list` of `string` features.
|
223 |
-
- `candidates`: a `list` of `string` features.
|
224 |
-
- `answer`: a `string` feature.
|
225 |
-
- `id`: a `string` feature.
|
226 |
-
|
227 |
-
### Data Splits
|
228 |
-
|
229 |
-
| name |train|validation|
|
230 |
-
|--------------|----:|---------:|
|
231 |
-
|masked_medhop | 1620| 342|
|
232 |
-
|masked_wikihop|43738| 5129|
|
233 |
-
|medhop | 1620| 342|
|
234 |
-
|wikihop |43738| 5129|
|
235 |
-
|
236 |
-
## Dataset Creation
|
237 |
-
|
238 |
-
### Curation Rationale
|
239 |
-
|
240 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
241 |
-
|
242 |
-
### Source Data
|
243 |
-
|
244 |
-
#### Initial Data Collection and Normalization
|
245 |
-
|
246 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
247 |
-
|
248 |
-
#### Who are the source language producers?
|
249 |
-
|
250 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
251 |
-
|
252 |
-
### Annotations
|
253 |
-
|
254 |
-
#### Annotation process
|
255 |
-
|
256 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
257 |
-
|
258 |
-
#### Who are the annotators?
|
259 |
-
|
260 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
261 |
-
|
262 |
-
### Personal and Sensitive Information
|
263 |
-
|
264 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
265 |
-
|
266 |
-
## Considerations for Using the Data
|
267 |
-
|
268 |
-
### Social Impact of Dataset
|
269 |
-
|
270 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
271 |
-
|
272 |
-
### Discussion of Biases
|
273 |
-
|
274 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
275 |
-
|
276 |
-
### Other Known Limitations
|
277 |
-
|
278 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
279 |
-
|
280 |
-
## Additional Information
|
281 |
-
|
282 |
-
### Dataset Curators
|
283 |
-
|
284 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
285 |
-
|
286 |
-
### Licensing Information
|
287 |
-
|
288 |
-
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
289 |
-
|
290 |
-
### Citation Information
|
291 |
-
|
292 |
-
```
|
293 |
-
|
294 |
-
```
|
295 |
-
|
296 |
-
|
297 |
-
### Contributions
|
298 |
-
|
299 |
-
Thanks to [@thomwolf](https://github.com/thomwolf), [@jplu](https://github.com/jplu), [@lewtun](https://github.com/lewtun), [@lhoestq](https://github.com/lhoestq), [@mariamabarham](https://github.com/mariamabarham) for adding this dataset.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
dataset_infos.json
DELETED
@@ -1 +0,0 @@
|
|
1 |
-
{"medhop": {"description": " We have created two new Reading Comprehension datasets focussing on multi-hop (alias multi-step) inference.\n\nSeveral pieces of information often jointly imply another fact. In multi-hop inference, a new fact is derived by combining facts via a chain of multiple steps.\n\nOur aim is to build Reading Comprehension methods that perform multi-hop inference on text, where individual facts are spread out across different documents.\n\nThe two QAngaroo datasets provide a training and evaluation resource for such methods.\n", "citation": "\n", "homepage": "http://qangaroo.cs.ucl.ac.uk/index.html", "license": "", "features": {"query": {"dtype": "string", "id": null, "_type": "Value"}, "supports": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "candidates": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "id": {"dtype": "string", "id": null, "_type": "Value"}}, "supervised_keys": null, "builder_name": "qangaroo", "config_name": "medhop", "version": {"version_str": "1.0.0", "description": "", "datasets_version_to_prepare": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 93947725, "num_examples": 1620, "dataset_name": "qangaroo"}, "validation": {"name": "validation", "num_bytes": 16463555, "num_examples": 342, "dataset_name": "qangaroo"}}, "download_checksums": {"https://drive.google.com/uc?export=download&id=1ytVZ4AhubFDOEL7o7XrIRIyhU8g9wvKA": {"num_bytes": 339843061, "checksum": "2f512869760cdad76a022a1465f025b486ae79dc5b8f0bf3ad901a4caf2d3050"}}, "download_size": 339843061, "dataset_size": 110411280, "size_in_bytes": 450254341}, "masked_medhop": {"description": " We have created two new Reading Comprehension datasets focussing on multi-hop (alias multi-step) inference.\n\nSeveral pieces of information often jointly imply another fact. In multi-hop inference, a new fact is derived by combining facts via a chain of multiple steps.\n\nOur aim is to build Reading Comprehension methods that perform multi-hop inference on text, where individual facts are spread out across different documents.\n\nThe two QAngaroo datasets provide a training and evaluation resource for such methods.\n", "citation": "\n", "homepage": "http://qangaroo.cs.ucl.ac.uk/index.html", "license": "", "features": {"query": {"dtype": "string", "id": null, "_type": "Value"}, "supports": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "candidates": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "id": {"dtype": "string", "id": null, "_type": "Value"}}, "supervised_keys": null, "builder_name": "qangaroo", "config_name": "masked_medhop", "version": {"version_str": "1.0.0", "description": "", "datasets_version_to_prepare": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 95823986, "num_examples": 1620, "dataset_name": "qangaroo"}, "validation": {"name": "validation", "num_bytes": 16802484, "num_examples": 342, "dataset_name": "qangaroo"}}, "download_checksums": {"https://drive.google.com/uc?export=download&id=1ytVZ4AhubFDOEL7o7XrIRIyhU8g9wvKA": {"num_bytes": 339843061, "checksum": "2f512869760cdad76a022a1465f025b486ae79dc5b8f0bf3ad901a4caf2d3050"}}, "download_size": 339843061, "dataset_size": 112626470, "size_in_bytes": 452469531}, "wikihop": {"description": " We have created two new Reading Comprehension datasets focussing on multi-hop (alias multi-step) inference.\n\nSeveral pieces of information often jointly imply another fact. In multi-hop inference, a new fact is derived by combining facts via a chain of multiple steps.\n\nOur aim is to build Reading Comprehension methods that perform multi-hop inference on text, where individual facts are spread out across different documents.\n\nThe two QAngaroo datasets provide a training and evaluation resource for such methods.\n", "citation": "\n", "homepage": "http://qangaroo.cs.ucl.ac.uk/index.html", "license": "", "features": {"query": {"dtype": "string", "id": null, "_type": "Value"}, "supports": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "candidates": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "id": {"dtype": "string", "id": null, "_type": "Value"}}, "supervised_keys": null, "builder_name": "qangaroo", "config_name": "wikihop", "version": {"version_str": "1.0.0", "description": "", "datasets_version_to_prepare": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 325994029, "num_examples": 43738, "dataset_name": "qangaroo"}, "validation": {"name": "validation", "num_bytes": 40869634, "num_examples": 5129, "dataset_name": "qangaroo"}}, "download_checksums": {"https://drive.google.com/uc?export=download&id=1ytVZ4AhubFDOEL7o7XrIRIyhU8g9wvKA": {"num_bytes": 339843061, "checksum": "2f512869760cdad76a022a1465f025b486ae79dc5b8f0bf3ad901a4caf2d3050"}}, "download_size": 339843061, "dataset_size": 366863663, "size_in_bytes": 706706724}, "masked_wikihop": {"description": " We have created two new Reading Comprehension datasets focussing on multi-hop (alias multi-step) inference.\n\nSeveral pieces of information often jointly imply another fact. In multi-hop inference, a new fact is derived by combining facts via a chain of multiple steps.\n\nOur aim is to build Reading Comprehension methods that perform multi-hop inference on text, where individual facts are spread out across different documents.\n\nThe two QAngaroo datasets provide a training and evaluation resource for such methods.\n", "citation": "\n", "homepage": "http://qangaroo.cs.ucl.ac.uk/index.html", "license": "", "features": {"query": {"dtype": "string", "id": null, "_type": "Value"}, "supports": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "candidates": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}, "id": {"dtype": "string", "id": null, "_type": "Value"}}, "supervised_keys": null, "builder_name": "qangaroo", "config_name": "masked_wikihop", "version": {"version_str": "1.0.0", "description": "", "datasets_version_to_prepare": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 348290479, "num_examples": 43738, "dataset_name": "qangaroo"}, "validation": {"name": "validation", "num_bytes": 43689810, "num_examples": 5129, "dataset_name": "qangaroo"}}, "download_checksums": {"https://drive.google.com/uc?export=download&id=1ytVZ4AhubFDOEL7o7XrIRIyhU8g9wvKA": {"num_bytes": 339843061, "checksum": "2f512869760cdad76a022a1465f025b486ae79dc5b8f0bf3ad901a4caf2d3050"}}, "download_size": 339843061, "dataset_size": 391980289, "size_in_bytes": 731823350}}
|
|
|
|
masked_medhop/qangaroo-train.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0fc78c9efde7e04dd1e51971fa46206c4af0b4562afe93ccc6821843e6383134
|
3 |
+
size 50027281
|
masked_medhop/qangaroo-validation.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bcad8391cee468ea1fd01c421b12aa56cb3b8b8bd15036cbc3a1d95598d88033
|
3 |
+
size 8774398
|
masked_wikihop/qangaroo-train.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9483f7454f2a090890a98c425c60300758e77ea7a288ffcb1d558bec6251d468
|
3 |
+
size 187715934
|
masked_wikihop/qangaroo-validation.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7eb5d8bf876bdf6b529807189988a461767e68af6f9e028f1b234f6c4fbcf3f6
|
3 |
+
size 23586547
|
medhop/qangaroo-train.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:377d3aacb4ffa3455b4f427041b7d430159d87d22f98a64be938e0ac2b19271d
|
3 |
+
size 49190500
|
medhop/qangaroo-validation.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e25d92075881443ef9d219aba6062141d9825979150fff1cdab7cd06e475e48e
|
3 |
+
size 8647216
|
qangaroo.py
DELETED
@@ -1,126 +0,0 @@
|
|
1 |
-
"""TODO(qangaroo): Add a description here."""
|
2 |
-
|
3 |
-
|
4 |
-
import json
|
5 |
-
import os
|
6 |
-
|
7 |
-
import datasets
|
8 |
-
|
9 |
-
|
10 |
-
# TODO(qangaroo): BibTeX citation
|
11 |
-
|
12 |
-
_CITATION = """
|
13 |
-
"""
|
14 |
-
|
15 |
-
# TODO(quangaroo):
|
16 |
-
_DESCRIPTION = """\
|
17 |
-
We have created two new Reading Comprehension datasets focussing on multi-hop (alias multi-step) inference.
|
18 |
-
|
19 |
-
Several pieces of information often jointly imply another fact. In multi-hop inference, a new fact is derived by combining facts via a chain of multiple steps.
|
20 |
-
|
21 |
-
Our aim is to build Reading Comprehension methods that perform multi-hop inference on text, where individual facts are spread out across different documents.
|
22 |
-
|
23 |
-
The two QAngaroo datasets provide a training and evaluation resource for such methods.
|
24 |
-
"""
|
25 |
-
|
26 |
-
_MEDHOP_DESCRIPTION = """\
|
27 |
-
With the same format as WikiHop, this dataset is based on research paper abstracts from PubMed, and the queries are about interactions between pairs of drugs.
|
28 |
-
The correct answer has to be inferred by combining information from a chain of reactions of drugs and proteins.
|
29 |
-
"""
|
30 |
-
_WIKIHOP_DESCRIPTION = """\
|
31 |
-
With the same format as WikiHop, this dataset is based on research paper abstracts from PubMed, and the queries are about interactions between pairs of drugs.
|
32 |
-
The correct answer has to be inferred by combining information from a chain of reactions of drugs and proteins.
|
33 |
-
"""
|
34 |
-
|
35 |
-
_URL = "https://drive.google.com/uc?export=download&id=1ytVZ4AhubFDOEL7o7XrIRIyhU8g9wvKA"
|
36 |
-
|
37 |
-
|
38 |
-
class QangarooConfig(datasets.BuilderConfig):
|
39 |
-
def __init__(self, data_dir, **kwargs):
|
40 |
-
"""BuilderConfig for qangaroo dataset
|
41 |
-
|
42 |
-
Args:
|
43 |
-
data_dir: directory for the given dataset name
|
44 |
-
**kwargs: keyword arguments forwarded to super.
|
45 |
-
|
46 |
-
"""
|
47 |
-
|
48 |
-
super(QangarooConfig, self).__init__(version=datasets.Version("1.0.0", ""), **kwargs)
|
49 |
-
|
50 |
-
self.data_dir = data_dir
|
51 |
-
|
52 |
-
|
53 |
-
class Qangaroo(datasets.GeneratorBasedBuilder):
|
54 |
-
"""TODO(qangaroo): Short description of my dataset."""
|
55 |
-
|
56 |
-
# TODO(qangaroo): Set up version.
|
57 |
-
VERSION = datasets.Version("0.1.0")
|
58 |
-
BUILDER_CONFIGS = [
|
59 |
-
QangarooConfig(name="medhop", description=_MEDHOP_DESCRIPTION, data_dir="medhop"),
|
60 |
-
QangarooConfig(name="masked_medhop", description=_MEDHOP_DESCRIPTION, data_dir="medhop"),
|
61 |
-
QangarooConfig(name="wikihop", description=_WIKIHOP_DESCRIPTION, data_dir="wikihop"),
|
62 |
-
QangarooConfig(name="masked_wikihop", description=_WIKIHOP_DESCRIPTION, data_dir="wikihop"),
|
63 |
-
]
|
64 |
-
|
65 |
-
def _info(self):
|
66 |
-
# TODO(qangaroo): Specifies the datasets.DatasetInfo object
|
67 |
-
return datasets.DatasetInfo(
|
68 |
-
# This is the description that will appear on the datasets page.
|
69 |
-
description=_DESCRIPTION,
|
70 |
-
# datasets.features.FeatureConnectors
|
71 |
-
features=datasets.Features(
|
72 |
-
{
|
73 |
-
# These are the features of your dataset like images, labels ...
|
74 |
-
"query": datasets.Value("string"),
|
75 |
-
"supports": datasets.features.Sequence(datasets.Value("string")),
|
76 |
-
"candidates": datasets.features.Sequence(datasets.Value("string")),
|
77 |
-
"answer": datasets.Value("string"),
|
78 |
-
"id": datasets.Value("string")
|
79 |
-
# These are the features of your dataset like images, labels ...
|
80 |
-
}
|
81 |
-
),
|
82 |
-
# If there's a common (input, target) tuple from the features,
|
83 |
-
# specify them here. They'll be used if as_supervised=True in
|
84 |
-
# builder.as_dataset.
|
85 |
-
supervised_keys=None,
|
86 |
-
# Homepage of the dataset for documentation
|
87 |
-
homepage="http://qangaroo.cs.ucl.ac.uk/index.html",
|
88 |
-
citation=_CITATION,
|
89 |
-
)
|
90 |
-
|
91 |
-
def _split_generators(self, dl_manager):
|
92 |
-
"""Returns SplitGenerators."""
|
93 |
-
# TODO(qangaroo): Downloads the data and defines the splits
|
94 |
-
# dl_manager is a datasets.download.DownloadManager that can be used to
|
95 |
-
# download and extract URLs
|
96 |
-
dl_dir = dl_manager.download_and_extract(_URL)
|
97 |
-
data_dir = os.path.join(dl_dir, "qangaroo_v1.1")
|
98 |
-
train_file = "train.masked.json" if "masked" in self.config.name else "train.json"
|
99 |
-
dev_file = "dev.masked.json" if "masked" in self.config.name else "dev.json"
|
100 |
-
return [
|
101 |
-
datasets.SplitGenerator(
|
102 |
-
name=datasets.Split.TRAIN,
|
103 |
-
# These kwargs will be passed to _generate_examples
|
104 |
-
gen_kwargs={"filepath": os.path.join(data_dir, self.config.data_dir, train_file)},
|
105 |
-
),
|
106 |
-
datasets.SplitGenerator(
|
107 |
-
name=datasets.Split.VALIDATION,
|
108 |
-
# These kwargs will be passed to _generate_examples
|
109 |
-
gen_kwargs={"filepath": os.path.join(data_dir, self.config.data_dir, dev_file)},
|
110 |
-
),
|
111 |
-
]
|
112 |
-
|
113 |
-
def _generate_examples(self, filepath):
|
114 |
-
"""Yields examples."""
|
115 |
-
# TODO(quangaroo): Yields (key, example) tuples from the dataset
|
116 |
-
with open(filepath, encoding="utf-8") as f:
|
117 |
-
data = json.load(f)
|
118 |
-
for example in data:
|
119 |
-
id_ = example["id"]
|
120 |
-
yield id_, {
|
121 |
-
"id": example["id"],
|
122 |
-
"query": example["query"],
|
123 |
-
"supports": example["supports"],
|
124 |
-
"candidates": example["candidates"],
|
125 |
-
"answer": example["answer"],
|
126 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
wikihop/qangaroo-train.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:779c4627f6db8ece7bfe2828562725372ad0d09b8375b07cd2fc07b24a18f543
|
3 |
+
size 179875980
|
wikihop/qangaroo-validation.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ae702f97333d4d81035543074f1743bf775bb324f38400753bd720d9968e18c1
|
3 |
+
size 22578466
|