Datasets:
Upload dataloading script and README.md
Browse files- README.md +565 -0
- knowledge_net.py +368 -0
README.md
ADDED
|
@@ -0,0 +1,565 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
annotations_creators:
|
| 3 |
+
- expert-generated
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
language_creators:
|
| 7 |
+
- found
|
| 8 |
+
license: []
|
| 9 |
+
multilinguality:
|
| 10 |
+
- monolingual
|
| 11 |
+
pretty_name: KnowledgeNet is a dataset for automatically populating a knowledge base
|
| 12 |
+
size_categories:
|
| 13 |
+
- 10K<n<100K
|
| 14 |
+
source_datasets: []
|
| 15 |
+
tags:
|
| 16 |
+
- knowledgenet
|
| 17 |
+
task_categories:
|
| 18 |
+
- text-classification
|
| 19 |
+
task_ids:
|
| 20 |
+
- multi-class-classification
|
| 21 |
+
- entity-linking-classification
|
| 22 |
+
dataset_info:
|
| 23 |
+
- config_name: knet
|
| 24 |
+
features:
|
| 25 |
+
- name: fold
|
| 26 |
+
dtype: int32
|
| 27 |
+
- name: documentId
|
| 28 |
+
dtype: string
|
| 29 |
+
- name: source
|
| 30 |
+
dtype: string
|
| 31 |
+
- name: documentText
|
| 32 |
+
dtype: string
|
| 33 |
+
- name: passages
|
| 34 |
+
sequence:
|
| 35 |
+
- name: passageId
|
| 36 |
+
dtype: string
|
| 37 |
+
- name: passageStart
|
| 38 |
+
dtype: int32
|
| 39 |
+
- name: passageEnd
|
| 40 |
+
dtype: int32
|
| 41 |
+
- name: passageText
|
| 42 |
+
dtype: string
|
| 43 |
+
- name: exhaustivelyAnnotatedProperties
|
| 44 |
+
sequence:
|
| 45 |
+
- name: propertyId
|
| 46 |
+
dtype: string
|
| 47 |
+
- name: propertyName
|
| 48 |
+
dtype: string
|
| 49 |
+
- name: propertyDescription
|
| 50 |
+
dtype: string
|
| 51 |
+
- name: facts
|
| 52 |
+
sequence:
|
| 53 |
+
- name: factId
|
| 54 |
+
dtype: string
|
| 55 |
+
- name: propertyId
|
| 56 |
+
dtype: string
|
| 57 |
+
- name: humanReadable
|
| 58 |
+
dtype: string
|
| 59 |
+
- name: annotatedPassage
|
| 60 |
+
dtype: string
|
| 61 |
+
- name: subjectStart
|
| 62 |
+
dtype: int32
|
| 63 |
+
- name: subjectEnd
|
| 64 |
+
dtype: int32
|
| 65 |
+
- name: subjectText
|
| 66 |
+
dtype: string
|
| 67 |
+
- name: subjectUri
|
| 68 |
+
dtype: string
|
| 69 |
+
- name: objectStart
|
| 70 |
+
dtype: int32
|
| 71 |
+
- name: objectEnd
|
| 72 |
+
dtype: int32
|
| 73 |
+
- name: objectText
|
| 74 |
+
dtype: string
|
| 75 |
+
- name: objectUri
|
| 76 |
+
dtype: string
|
| 77 |
+
splits:
|
| 78 |
+
- name: train
|
| 79 |
+
num_bytes: 10161415
|
| 80 |
+
num_examples: 3977
|
| 81 |
+
download_size: 14119313
|
| 82 |
+
dataset_size: 10161415
|
| 83 |
+
- config_name: knet_tokenized
|
| 84 |
+
features:
|
| 85 |
+
- name: doc_id
|
| 86 |
+
dtype: string
|
| 87 |
+
- name: passage_id
|
| 88 |
+
dtype: string
|
| 89 |
+
- name: fact_id
|
| 90 |
+
dtype: string
|
| 91 |
+
- name: tokens
|
| 92 |
+
sequence: string
|
| 93 |
+
- name: subj_start
|
| 94 |
+
dtype: int32
|
| 95 |
+
- name: subj_end
|
| 96 |
+
dtype: int32
|
| 97 |
+
- name: subj_type
|
| 98 |
+
dtype:
|
| 99 |
+
class_label:
|
| 100 |
+
names:
|
| 101 |
+
'0': O
|
| 102 |
+
'1': PER
|
| 103 |
+
'2': ORG
|
| 104 |
+
'3': LOC
|
| 105 |
+
'4': DATE
|
| 106 |
+
- name: subj_uri
|
| 107 |
+
dtype: string
|
| 108 |
+
- name: obj_start
|
| 109 |
+
dtype: int32
|
| 110 |
+
- name: obj_end
|
| 111 |
+
dtype: int32
|
| 112 |
+
- name: obj_type
|
| 113 |
+
dtype:
|
| 114 |
+
class_label:
|
| 115 |
+
names:
|
| 116 |
+
'0': O
|
| 117 |
+
'1': PER
|
| 118 |
+
'2': ORG
|
| 119 |
+
'3': LOC
|
| 120 |
+
'4': DATE
|
| 121 |
+
- name: obj_uri
|
| 122 |
+
dtype: string
|
| 123 |
+
- name: relation
|
| 124 |
+
dtype:
|
| 125 |
+
class_label:
|
| 126 |
+
names:
|
| 127 |
+
'0': NO_RELATION
|
| 128 |
+
'1': DATE_OF_BIRTH
|
| 129 |
+
'2': DATE_OF_DEATH
|
| 130 |
+
'3': PLACE_OF_RESIDENCE
|
| 131 |
+
'4': PLACE_OF_BIRTH
|
| 132 |
+
'5': NATIONALITY
|
| 133 |
+
'6': EMPLOYEE_OR_MEMBER_OF
|
| 134 |
+
'7': EDUCATED_AT
|
| 135 |
+
'8': POLITICAL_AFFILIATION
|
| 136 |
+
'9': CHILD_OF
|
| 137 |
+
'10': SPOUSE
|
| 138 |
+
'11': DATE_FOUNDED
|
| 139 |
+
'12': HEADQUARTERS
|
| 140 |
+
'13': SUBSIDIARY_OF
|
| 141 |
+
'14': FOUNDED_BY
|
| 142 |
+
'15': CEO
|
| 143 |
+
splits:
|
| 144 |
+
- name: train
|
| 145 |
+
num_bytes: 4511963
|
| 146 |
+
num_examples: 10895
|
| 147 |
+
download_size: 14119313
|
| 148 |
+
dataset_size: 4511963
|
| 149 |
+
- config_name: knet_re
|
| 150 |
+
features:
|
| 151 |
+
- name: documentId
|
| 152 |
+
dtype: string
|
| 153 |
+
- name: passageId
|
| 154 |
+
dtype: string
|
| 155 |
+
- name: factId
|
| 156 |
+
dtype: string
|
| 157 |
+
- name: passageText
|
| 158 |
+
dtype: string
|
| 159 |
+
- name: humanReadable
|
| 160 |
+
dtype: string
|
| 161 |
+
- name: annotatedPassage
|
| 162 |
+
dtype: string
|
| 163 |
+
- name: subjectStart
|
| 164 |
+
dtype: int32
|
| 165 |
+
- name: subjectEnd
|
| 166 |
+
dtype: int32
|
| 167 |
+
- name: subjectText
|
| 168 |
+
dtype: string
|
| 169 |
+
- name: subjectType
|
| 170 |
+
dtype:
|
| 171 |
+
class_label:
|
| 172 |
+
names:
|
| 173 |
+
'0': O
|
| 174 |
+
'1': PER
|
| 175 |
+
'2': ORG
|
| 176 |
+
'3': LOC
|
| 177 |
+
'4': DATE
|
| 178 |
+
- name: subjectUri
|
| 179 |
+
dtype: string
|
| 180 |
+
- name: objectStart
|
| 181 |
+
dtype: int32
|
| 182 |
+
- name: objectEnd
|
| 183 |
+
dtype: int32
|
| 184 |
+
- name: objectText
|
| 185 |
+
dtype: string
|
| 186 |
+
- name: objectType
|
| 187 |
+
dtype:
|
| 188 |
+
class_label:
|
| 189 |
+
names:
|
| 190 |
+
'0': O
|
| 191 |
+
'1': PER
|
| 192 |
+
'2': ORG
|
| 193 |
+
'3': LOC
|
| 194 |
+
'4': DATE
|
| 195 |
+
- name: objectUri
|
| 196 |
+
dtype: string
|
| 197 |
+
- name: relation
|
| 198 |
+
dtype:
|
| 199 |
+
class_label:
|
| 200 |
+
names:
|
| 201 |
+
'0': NO_RELATION
|
| 202 |
+
'1': DATE_OF_BIRTH
|
| 203 |
+
'2': DATE_OF_DEATH
|
| 204 |
+
'3': PLACE_OF_RESIDENCE
|
| 205 |
+
'4': PLACE_OF_BIRTH
|
| 206 |
+
'5': NATIONALITY
|
| 207 |
+
'6': EMPLOYEE_OR_MEMBER_OF
|
| 208 |
+
'7': EDUCATED_AT
|
| 209 |
+
'8': POLITICAL_AFFILIATION
|
| 210 |
+
'9': CHILD_OF
|
| 211 |
+
'10': SPOUSE
|
| 212 |
+
'11': DATE_FOUNDED
|
| 213 |
+
'12': HEADQUARTERS
|
| 214 |
+
'13': SUBSIDIARY_OF
|
| 215 |
+
'14': FOUNDED_BY
|
| 216 |
+
'15': CEO
|
| 217 |
+
splits:
|
| 218 |
+
- name: train
|
| 219 |
+
num_bytes: 6098219
|
| 220 |
+
num_examples: 10895
|
| 221 |
+
download_size: 14119313
|
| 222 |
+
dataset_size: 6098219
|
| 223 |
+
---
|
| 224 |
+
# Dataset Card for "KnowledgeNet"
|
| 225 |
+
## Table of Contents
|
| 226 |
+
- [Table of Contents](#table-of-contents)
|
| 227 |
+
- [Dataset Description](#dataset-description)
|
| 228 |
+
- [Dataset Summary](#dataset-summary)
|
| 229 |
+
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
|
| 230 |
+
- [Languages](#languages)
|
| 231 |
+
- [Dataset Structure](#dataset-structure)
|
| 232 |
+
- [Data Instances](#data-instances)
|
| 233 |
+
- [Data Fields](#data-fields)
|
| 234 |
+
- [Data Splits](#data-splits)
|
| 235 |
+
- [Dataset Creation](#dataset-creation)
|
| 236 |
+
- [Curation Rationale](#curation-rationale)
|
| 237 |
+
- [Source Data](#source-data)
|
| 238 |
+
- [Annotations](#annotations)
|
| 239 |
+
- [Personal and Sensitive Information](#personal-and-sensitive-information)
|
| 240 |
+
- [Considerations for Using the Data](#considerations-for-using-the-data)
|
| 241 |
+
- [Social Impact of Dataset](#social-impact-of-dataset)
|
| 242 |
+
- [Discussion of Biases](#discussion-of-biases)
|
| 243 |
+
- [Other Known Limitations](#other-known-limitations)
|
| 244 |
+
- [Additional Information](#additional-information)
|
| 245 |
+
- [Dataset Curators](#dataset-curators)
|
| 246 |
+
- [Licensing Information](#licensing-information)
|
| 247 |
+
- [Citation Information](#citation-information)
|
| 248 |
+
- [Contributions](#contributions)
|
| 249 |
+
## Dataset Description
|
| 250 |
+
- **Repository:** [knowledge-net](https://github.com/diffbot/knowledge-net)
|
| 251 |
+
- **Paper:** [KnowledgeNet: A Benchmark Dataset for Knowledge Base Population](https://aclanthology.org/D19-1069/)
|
| 252 |
+
- **Size of downloaded dataset files:** 12.59 MB
|
| 253 |
+
- **Size of the generated dataset:** 6.1 MB
|
| 254 |
+
### Dataset Summary
|
| 255 |
+
KnowledgeNet is a benchmark dataset for the task of automatically populating a knowledge base (Wikidata) with facts
|
| 256 |
+
expressed in natural language text on the web. KnowledgeNet provides text exhaustively annotated with facts, thus
|
| 257 |
+
enabling the holistic end-to-end evaluation of knowledge base population systems as a whole, unlike previous benchmarks
|
| 258 |
+
that are more suitable for the evaluation of individual subcomponents (e.g., entity linking, relation extraction).
|
| 259 |
+
|
| 260 |
+
For instance, the dataset contains text expressing the fact (Gennaro Basile; RESIDENCE; Moravia), in the passage:
|
| 261 |
+
"Gennaro Basile was an Italian painter, born in Naples but active in the German-speaking countries. He settled at Brünn,
|
| 262 |
+
in Moravia, and lived about 1756..."
|
| 263 |
+
|
| 264 |
+
For a description of the dataset and baseline systems, please refer to their
|
| 265 |
+
[EMNLP paper](https://github.com/diffbot/knowledge-net/blob/master/knowledgenet-emnlp-cameraready.pdf).
|
| 266 |
+
|
| 267 |
+
Note: This Datasetreader currently only supports the `train` split and does not contain negative examples.
|
| 268 |
+
In addition to the original format this repository also provides two version (`knet_re`, `knet_tokenized`) that are
|
| 269 |
+
easier to use for simple relation extraction. You can load them with
|
| 270 |
+
`datasets.load_dataset("DFKI-SLT/knowledge_net", name="<config>")`.
|
| 271 |
+
|
| 272 |
+
### Supported Tasks and Leaderboards
|
| 273 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 274 |
+
|
| 275 |
+
### Languages
|
| 276 |
+
The language in the dataset is English.
|
| 277 |
+
|
| 278 |
+
## Dataset Structure
|
| 279 |
+
### Data Instances
|
| 280 |
+
#### knet
|
| 281 |
+
- **Size of downloaded dataset files:** 12.59 MB
|
| 282 |
+
- **Size of the generated dataset:** 10.16 MB
|
| 283 |
+
|
| 284 |
+
An example of 'train' looks as follows:
|
| 285 |
+
```json
|
| 286 |
+
{
|
| 287 |
+
"fold": 2,
|
| 288 |
+
"documentId": "8313",
|
| 289 |
+
"source": "DBpedia Abstract",
|
| 290 |
+
"documentText": "Gennaro Basile\n\nGennaro Basile was an Italian painter, born in Naples but active in the German-speaking countries. He settled at Brünn, in Moravia, and lived about 1756. His best picture is the altar-piece in the chapel of the chateau at Seeberg, in Salzburg. Most of his works remained in Moravia.",
|
| 291 |
+
"passages": [
|
| 292 |
+
{
|
| 293 |
+
"passageId": "8313:16:114",
|
| 294 |
+
"passageStart": 16,
|
| 295 |
+
"passageEnd": 114,
|
| 296 |
+
"passageText": "Gennaro Basile was an Italian painter, born in Naples but active in the German-speaking countries.",
|
| 297 |
+
"exhaustivelyAnnotatedProperties": [
|
| 298 |
+
{
|
| 299 |
+
"propertyId": "12",
|
| 300 |
+
"propertyName": "PLACE_OF_BIRTH",
|
| 301 |
+
"propertyDescription": "Describes the relationship between a person and the location where she/he was born."
|
| 302 |
+
}
|
| 303 |
+
],
|
| 304 |
+
"facts": [
|
| 305 |
+
{
|
| 306 |
+
"factId": "8313:16:30:63:69:12",
|
| 307 |
+
"propertyId": "12",
|
| 308 |
+
"humanReadable": "<Gennaro Basile> <PLACE_OF_BIRTH> <Naples>",
|
| 309 |
+
"annotatedPassage": "<Gennaro Basile> was an Italian painter, born in <Naples> but active in the German-speaking countries.",
|
| 310 |
+
"subjectStart": 16,
|
| 311 |
+
"subjectEnd": 30,
|
| 312 |
+
"subjectText": "Gennaro Basile",
|
| 313 |
+
"subjectUri": "http://www.wikidata.org/entity/Q19517888",
|
| 314 |
+
"objectStart": 63,
|
| 315 |
+
"objectEnd": 69,
|
| 316 |
+
"objectText": "Naples",
|
| 317 |
+
"objectUri": "http://www.wikidata.org/entity/Q2634"
|
| 318 |
+
}
|
| 319 |
+
]
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"passageId": "8313:115:169",
|
| 323 |
+
"passageStart": 115,
|
| 324 |
+
"passageEnd": 169,
|
| 325 |
+
"passageText": "He settled at Brünn, in Moravia, and lived about 1756.",
|
| 326 |
+
"exhaustivelyAnnotatedProperties": [
|
| 327 |
+
{
|
| 328 |
+
"propertyId": "11",
|
| 329 |
+
"propertyName": "PLACE_OF_RESIDENCE",
|
| 330 |
+
"propertyDescription": "Describes the relationship between a person and the location where she/he lives/lived."
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"propertyId": "12",
|
| 334 |
+
"propertyName": "PLACE_OF_BIRTH",
|
| 335 |
+
"propertyDescription": "Describes the relationship between a person and the location where she/he was born."
|
| 336 |
+
}
|
| 337 |
+
],
|
| 338 |
+
"facts": [
|
| 339 |
+
{
|
| 340 |
+
"factId": "8313:115:117:129:134:11",
|
| 341 |
+
"propertyId": "11",
|
| 342 |
+
"humanReadable": "<He> <PLACE_OF_RESIDENCE> <Brünn>",
|
| 343 |
+
"annotatedPassage": "<He> settled at <Brünn>, in Moravia, and lived about 1756.",
|
| 344 |
+
"subjectStart": 115,
|
| 345 |
+
"subjectEnd": 117,
|
| 346 |
+
"subjectText": "He",
|
| 347 |
+
"subjectUri": "http://www.wikidata.org/entity/Q19517888",
|
| 348 |
+
"objectStart": 129,
|
| 349 |
+
"objectEnd": 134,
|
| 350 |
+
"objectText": "Brünn",
|
| 351 |
+
"objectUri": "http://www.wikidata.org/entity/Q14960"
|
| 352 |
+
},
|
| 353 |
+
{
|
| 354 |
+
"factId": "8313:115:117:139:146:11",
|
| 355 |
+
"propertyId": "11",
|
| 356 |
+
"humanReadable": "<He> <PLACE_OF_RESIDENCE> <Moravia>",
|
| 357 |
+
"annotatedPassage": "<He> settled at Brünn, in <Moravia>, and lived about 1756.",
|
| 358 |
+
"subjectStart": 115,
|
| 359 |
+
"subjectEnd": 117,
|
| 360 |
+
"subjectText": "He",
|
| 361 |
+
"subjectUri": "http://www.wikidata.org/entity/Q19517888",
|
| 362 |
+
"objectStart": 139,
|
| 363 |
+
"objectEnd": 146,
|
| 364 |
+
"objectText": "Moravia",
|
| 365 |
+
"objectUri": "http://www.wikidata.org/entity/Q43266"
|
| 366 |
+
}
|
| 367 |
+
]
|
| 368 |
+
}
|
| 369 |
+
]
|
| 370 |
+
}
|
| 371 |
+
```
|
| 372 |
+
|
| 373 |
+
#### knet_re
|
| 374 |
+
- **Size of downloaded dataset files:** 12.59 MB
|
| 375 |
+
- **Size of the generated dataset:** 6.1 MB
|
| 376 |
+
|
| 377 |
+
An example of 'train' looks as follows:
|
| 378 |
+
```json
|
| 379 |
+
{
|
| 380 |
+
"documentId": "7",
|
| 381 |
+
"passageId": "7:23:206",
|
| 382 |
+
"factId": "7:23:44:138:160:1",
|
| 383 |
+
"passageText": "Tata Chemicals Europe (formerly Brunner Mond (UK) Limited) is a UK-based chemicals company that is a subsidiary of Tata Chemicals Limited, itself a part of the India-based Tata Group.",
|
| 384 |
+
"humanReadable": "<Tata Chemicals Europe> <SUBSIDIARY_OF> <Tata Chemicals Limited>",
|
| 385 |
+
"annotatedPassage": "<Tata Chemicals Europe> (formerly Brunner Mond (UK) Limited) is a UK-based chemicals company that is a subsidiary of <Tata Chemicals Limited>, itself a part of the India-based Tata Group.",
|
| 386 |
+
"subjectStart": 0,
|
| 387 |
+
"subjectEnd": 21,
|
| 388 |
+
"subjectText": "Tata Chemicals Europe",
|
| 389 |
+
"subjectType": 2,
|
| 390 |
+
"subjectUri": "",
|
| 391 |
+
"objectStart": 115,
|
| 392 |
+
"objectEnd": 137,
|
| 393 |
+
"objectText": "Tata Chemicals Limited",
|
| 394 |
+
"objectType": 2,
|
| 395 |
+
"objectUri": "http://www.wikidata.org/entity/Q2331365",
|
| 396 |
+
"relation": 13
|
| 397 |
+
}
|
| 398 |
+
```
|
| 399 |
+
|
| 400 |
+
#### knet_tokenized
|
| 401 |
+
- **Size of downloaded dataset files:** 12.59 MB
|
| 402 |
+
- **Size of the generated dataset:** 4.5 MB
|
| 403 |
+
|
| 404 |
+
An example of 'train' looks as follows:
|
| 405 |
+
```json
|
| 406 |
+
{
|
| 407 |
+
"doc_id": "7",
|
| 408 |
+
"passage_id": "7:23:206",
|
| 409 |
+
"fact_id": "7:162:168:183:205:1",
|
| 410 |
+
"tokens": ["Tata", "Chemicals", "Europe", "(", "formerly", "Brunner", "Mond", "(", "UK", ")", "Limited", ")", "is", "a", "UK", "-", "based", "chemicals", "company", "that", "is", "a", "subsidiary", "of", "Tata", "Chemicals", "Limited", ",", "itself", "a", "part", "of", "the", "India", "-", "based", "Tata", "Group", "."],
|
| 411 |
+
"subj_start": 28,
|
| 412 |
+
"subj_end": 29,
|
| 413 |
+
"subj_type": 2,
|
| 414 |
+
"subj_uri": "http://www.wikidata.org/entity/Q2331365",
|
| 415 |
+
"obj_start": 33,
|
| 416 |
+
"obj_end": 38,
|
| 417 |
+
"obj_type": 2,
|
| 418 |
+
"obj_uri": "http://www.wikidata.org/entity/Q331715",
|
| 419 |
+
"relation": 13
|
| 420 |
+
}
|
| 421 |
+
```
|
| 422 |
+
### Data Fields
|
| 423 |
+
|
| 424 |
+
#### knet
|
| 425 |
+
- `fold`: the fold, a `int` feature.
|
| 426 |
+
- `documentId`: the document id, a `string` feature.
|
| 427 |
+
- `source`: the source, a `string` feature.
|
| 428 |
+
- `documenText`: the document text, a `string` feature.
|
| 429 |
+
- `passages`: the list of passages, a `list` of `dict`.
|
| 430 |
+
- `passageId`: the passage id, a `string` feature.
|
| 431 |
+
- `passageStart`: the passage start, a `int` feature.
|
| 432 |
+
- `passageEnd`: the passage end, a `int` feature.
|
| 433 |
+
- `passageText`: the passage text, a `string` feature.
|
| 434 |
+
- `exhaustivelyAnnotatedProperties`: the list of exhaustively annotated properties, a `list` of `dict`.
|
| 435 |
+
- `propertyId`: the property id, a `string` feature.
|
| 436 |
+
- `propertyName`: the property name, a `string` feature.
|
| 437 |
+
- `propertyDescription`: the property description, a `string` feature.
|
| 438 |
+
- `facts`: the list of facts, a `list` of `dict`.
|
| 439 |
+
- `factId`: the fact id, a `string` feature.
|
| 440 |
+
- `propertyId`: the property id, a `string` feature.
|
| 441 |
+
- `humanReadable`: the human readable annotation, a `string` feature.
|
| 442 |
+
- `annotatedPassage`: the annotated passage, a `string` feature.
|
| 443 |
+
- `subjectStart`: the subject start, a `int` feature.
|
| 444 |
+
- `subjectEnd`: the subject end, a `int` feature.
|
| 445 |
+
- `subjectText`: the subject text, a `string` feature.
|
| 446 |
+
- `subjectUri`: the subject uri, a `string` feature.
|
| 447 |
+
- `objectStart`: the object start, a `int` feature.
|
| 448 |
+
- `objectEnd`: the object end, a `int` feature.
|
| 449 |
+
- `objectText`: the object text, a `string` feature.
|
| 450 |
+
- `objectUri`: the object uri, a `string` feature.
|
| 451 |
+
|
| 452 |
+
#### knet_re
|
| 453 |
+
- `documentId`: the document id, a `string` feature.
|
| 454 |
+
- `passageId`: the passage id, a `string` feature.
|
| 455 |
+
- `passageText`: the passage text, a `string` feature.
|
| 456 |
+
- `factId`: the fact id, a `string` feature.
|
| 457 |
+
- `humanReadable`: human-readable annotation, a `string` features.
|
| 458 |
+
- `annotatedPassage`: annotated passage, a `string` feature.
|
| 459 |
+
- `subjectStart`: the index of the start character of the relation subject mention, an `ìnt` feature.
|
| 460 |
+
- `subjectEnd`: the index of the end character of the relation subject mention, exclusive, an `ìnt` feature.
|
| 461 |
+
- `subjectText`: the text the subject mention, a `string` feature.
|
| 462 |
+
- `subjectType`: the NER type of the subject mention, a `string` classification label.
|
| 463 |
+
|
| 464 |
+
```json
|
| 465 |
+
{"O": 0, "PER": 1, "ORG": 2, "LOC": 3, "DATE": 4}
|
| 466 |
+
```
|
| 467 |
+
|
| 468 |
+
- `subjectUri`: the Wikidata URI of the subject mention, a `string` feature.
|
| 469 |
+
- `objectStart`: the index of the start character of the relation object mention, an `ìnt` feature.
|
| 470 |
+
- `objectEnd`: the index of the end character of the relation object mention, exclusive, an `ìnt` feature.
|
| 471 |
+
- `objectText`: the text the object mention, a `string` feature.
|
| 472 |
+
- `objectType`: the NER type of the object mention, a `string` classification label.
|
| 473 |
+
|
| 474 |
+
```json
|
| 475 |
+
{"O": 0, "PER": 1, "ORG": 2, "LOC": 3, "DATE": 4}
|
| 476 |
+
```
|
| 477 |
+
|
| 478 |
+
- `objectUri`: the Wikidata URI of the object mention, a `string` feature.
|
| 479 |
+
- `relation`: the relation label of this instance, a `string` classification label.
|
| 480 |
+
|
| 481 |
+
```json
|
| 482 |
+
{"NO_RELATION": 0, "DATE_OF_BIRTH": 1, "DATE_OF_DEATH": 2, "PLACE_OF_RESIDENCE": 3, "PLACE_OF_BIRTH": 4, "NATIONALITY": 5, "EMPLOYEE_OR_MEMBER_OF": 6, "EDUCATED_AT": 7, "POLITICAL_AFFILIATION": 8, "CHILD_OF": 9, "SPOUSE": 10, "DATE_FOUNDED": 11, "HEADQUARTERS": 12, "SUBSIDIARY_OF": 13, "FOUNDED_BY": 14, "CEO": 15}
|
| 483 |
+
```
|
| 484 |
+
|
| 485 |
+
#### knet_tokenized
|
| 486 |
+
- `doc_id`: the document id, a `string` feature.
|
| 487 |
+
- `passage_id`: the passage id, a `string` feature.
|
| 488 |
+
- `factId`: the fact id, a `string` feature.
|
| 489 |
+
- `tokens`: the list of tokens of this passage, obtained with spaCy, a `list` of `string` features.
|
| 490 |
+
- `subj_start`: the index of the start token of the relation subject mention, an `ìnt` feature.
|
| 491 |
+
- `subj_end`: the index of the end token of the relation subject mention, exclusive, an `ìnt` feature.
|
| 492 |
+
- `subj_type`: the NER type of the subject mention, a `string` classification label.
|
| 493 |
+
|
| 494 |
+
```json
|
| 495 |
+
{"O": 0, "PER": 1, "ORG": 2, "LOC": 3, "DATE": 4}
|
| 496 |
+
```
|
| 497 |
+
|
| 498 |
+
|
| 499 |
+
- `subj_uri`: the Wikidata URI of the subject mention, a `string` feature.
|
| 500 |
+
- `obj_start`: the index of the start token of the relation object mention, an `ìnt` feature.
|
| 501 |
+
- `obj_end`: the index of the end token of the relation object mention, exclusive, an `ìnt` feature.
|
| 502 |
+
- `obj_type`: the NER type of the object mention, a `string` classification label.
|
| 503 |
+
|
| 504 |
+
```json
|
| 505 |
+
{"O": 0, "PER": 1, "ORG": 2, "LOC": 3, "DATE": 4}
|
| 506 |
+
```
|
| 507 |
+
|
| 508 |
+
- `obj_uri`: the Wikidata URI of the object mention, a `string` feature.
|
| 509 |
+
- `relation`: the relation label of this instance, a `string` classification label.
|
| 510 |
+
|
| 511 |
+
```json
|
| 512 |
+
{"NO_RELATION": 0, "DATE_OF_BIRTH": 1, "DATE_OF_DEATH": 2, "PLACE_OF_RESIDENCE": 3, "PLACE_OF_BIRTH": 4, "NATIONALITY": 5, "EMPLOYEE_OR_MEMBER_OF": 6, "EDUCATED_AT": 7, "POLITICAL_AFFILIATION": 8, "CHILD_OF": 9, "SPOUSE": 10, "DATE_FOUNDED": 11, "HEADQUARTERS": 12, "SUBSIDIARY_OF": 13, "FOUNDED_BY": 14, "CEO": 15}
|
| 513 |
+
```
|
| 514 |
+
|
| 515 |
+
|
| 516 |
+
### Data Splits
|
| 517 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 518 |
+
## Dataset Creation
|
| 519 |
+
### Curation Rationale
|
| 520 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 521 |
+
### Source Data
|
| 522 |
+
#### Initial Data Collection and Normalization
|
| 523 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 524 |
+
#### Who are the source language producers?
|
| 525 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 526 |
+
### Annotations
|
| 527 |
+
#### Annotation process
|
| 528 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 529 |
+
are labeled as no_relation.
|
| 530 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 531 |
+
### Personal and Sensitive Information
|
| 532 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 533 |
+
## Considerations for Using the Data
|
| 534 |
+
### Social Impact of Dataset
|
| 535 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 536 |
+
### Discussion of Biases
|
| 537 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 538 |
+
### Other Known Limitations
|
| 539 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 540 |
+
## Additional Information
|
| 541 |
+
### Dataset Curators
|
| 542 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 543 |
+
### Licensing Information
|
| 544 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
| 545 |
+
### Citation Information
|
| 546 |
+
```
|
| 547 |
+
@inproceedings{mesquita-etal-2019-knowledgenet,
|
| 548 |
+
title = "{K}nowledge{N}et: A Benchmark Dataset for Knowledge Base Population",
|
| 549 |
+
author = "Mesquita, Filipe and
|
| 550 |
+
Cannaviccio, Matteo and
|
| 551 |
+
Schmidek, Jordan and
|
| 552 |
+
Mirza, Paramita and
|
| 553 |
+
Barbosa, Denilson",
|
| 554 |
+
booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
|
| 555 |
+
month = nov,
|
| 556 |
+
year = "2019",
|
| 557 |
+
address = "Hong Kong, China",
|
| 558 |
+
publisher = "Association for Computational Linguistics",
|
| 559 |
+
url = "https://aclanthology.org/D19-1069",
|
| 560 |
+
doi = "10.18653/v1/D19-1069",
|
| 561 |
+
pages = "749--758",}
|
| 562 |
+
```
|
| 563 |
+
|
| 564 |
+
### Contributions
|
| 565 |
+
Thanks to [@phucdev](https://github.com/phucdev) for adding this dataset.
|
knowledge_net.py
ADDED
|
@@ -0,0 +1,368 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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 KnowledgeNet dataset for automatically populating a knowledge base"""
|
| 17 |
+
|
| 18 |
+
import json
|
| 19 |
+
import re
|
| 20 |
+
import datasets
|
| 21 |
+
|
| 22 |
+
_CITATION = """\
|
| 23 |
+
@inproceedings{mesquita-etal-2019-knowledgenet,
|
| 24 |
+
title = "{K}nowledge{N}et: A Benchmark Dataset for Knowledge Base Population",
|
| 25 |
+
author = "Mesquita, Filipe and
|
| 26 |
+
Cannaviccio, Matteo and
|
| 27 |
+
Schmidek, Jordan and
|
| 28 |
+
Mirza, Paramita and
|
| 29 |
+
Barbosa, Denilson",
|
| 30 |
+
booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
|
| 31 |
+
month = nov,
|
| 32 |
+
year = "2019",
|
| 33 |
+
address = "Hong Kong, China",
|
| 34 |
+
publisher = "Association for Computational Linguistics",
|
| 35 |
+
url = "https://aclanthology.org/D19-1069",
|
| 36 |
+
doi = "10.18653/v1/D19-1069",
|
| 37 |
+
pages = "749--758",}
|
| 38 |
+
"""
|
| 39 |
+
|
| 40 |
+
_DESCRIPTION = """\
|
| 41 |
+
KnowledgeNet is a benchmark dataset for the task of automatically populating a knowledge base (Wikidata) with facts
|
| 42 |
+
expressed in natural language text on the web. KnowledgeNet provides text exhaustively annotated with facts, thus
|
| 43 |
+
enabling the holistic end-to-end evaluation of knowledge base population systems as a whole, unlike previous benchmarks
|
| 44 |
+
that are more suitable for the evaluation of individual subcomponents (e.g., entity linking, relation extraction).
|
| 45 |
+
|
| 46 |
+
For instance, the dataset contains text expressing the fact (Gennaro Basile; RESIDENCE; Moravia), in the passage:
|
| 47 |
+
"Gennaro Basile was an Italian painter, born in Naples but active in the German-speaking countries. He settled at Brünn,
|
| 48 |
+
in Moravia, and lived about 1756..."
|
| 49 |
+
|
| 50 |
+
For a description of the dataset and baseline systems, please refer to their
|
| 51 |
+
[EMNLP paper](https://github.com/diffbot/knowledge-net/blob/master/knowledgenet-emnlp-cameraready.pdf).
|
| 52 |
+
|
| 53 |
+
Note: This Datasetreader currently only supports the `train` split and does not contain negative examples
|
| 54 |
+
"""
|
| 55 |
+
|
| 56 |
+
_HOMEPAGE = "https://github.com/diffbot/knowledge-net"
|
| 57 |
+
|
| 58 |
+
_LICENSE = ""
|
| 59 |
+
|
| 60 |
+
# The HuggingFace dataset library don't host the datasets but only point to the original files
|
| 61 |
+
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
| 62 |
+
_URLS = {
|
| 63 |
+
"train": "https://raw.githubusercontent.com/diffbot/knowledge-net/master/dataset/train.json",
|
| 64 |
+
"test": "https://raw.githubusercontent.com/diffbot/knowledge-net/master/dataset/test-no-facts.json"
|
| 65 |
+
}
|
| 66 |
+
|
| 67 |
+
_VERSION = datasets.Version("1.1.0")
|
| 68 |
+
|
| 69 |
+
_CLASS_LABELS = [
|
| 70 |
+
"NO_RELATION",
|
| 71 |
+
"DATE_OF_BIRTH",
|
| 72 |
+
"DATE_OF_DEATH",
|
| 73 |
+
"PLACE_OF_RESIDENCE",
|
| 74 |
+
"PLACE_OF_BIRTH",
|
| 75 |
+
"NATIONALITY",
|
| 76 |
+
"EMPLOYEE_OR_MEMBER_OF",
|
| 77 |
+
"EDUCATED_AT",
|
| 78 |
+
"POLITICAL_AFFILIATION",
|
| 79 |
+
"CHILD_OF",
|
| 80 |
+
"SPOUSE",
|
| 81 |
+
"DATE_FOUNDED",
|
| 82 |
+
"HEADQUARTERS",
|
| 83 |
+
"SUBSIDIARY_OF",
|
| 84 |
+
"FOUNDED_BY",
|
| 85 |
+
"CEO"
|
| 86 |
+
]
|
| 87 |
+
|
| 88 |
+
_NER_CLASS_LABELS = [
|
| 89 |
+
"O",
|
| 90 |
+
"PER",
|
| 91 |
+
"ORG",
|
| 92 |
+
"LOC",
|
| 93 |
+
"DATE"
|
| 94 |
+
]
|
| 95 |
+
|
| 96 |
+
|
| 97 |
+
def get_entity_types_from_relation(relation_label):
|
| 98 |
+
if relation_label == "DATE_OF_BIRTH":
|
| 99 |
+
subj_type = "PER"
|
| 100 |
+
obj_type = "DATE"
|
| 101 |
+
elif relation_label == "DATE_OF_DEATH":
|
| 102 |
+
subj_type = "PER"
|
| 103 |
+
obj_type = "DATE"
|
| 104 |
+
elif relation_label == "PLACE_OF_RESIDENCE":
|
| 105 |
+
subj_type = "PER"
|
| 106 |
+
obj_type = "LOC"
|
| 107 |
+
elif relation_label == "PLACE_OF_BIRTH":
|
| 108 |
+
subj_type = "PER"
|
| 109 |
+
obj_type = "LOC"
|
| 110 |
+
elif relation_label == "NATIONALITY":
|
| 111 |
+
subj_type = "PER"
|
| 112 |
+
obj_type = "LOC"
|
| 113 |
+
elif relation_label == "EMPLOYEE_OR_MEMBER_OF":
|
| 114 |
+
subj_type = "PER"
|
| 115 |
+
obj_type = "ORG"
|
| 116 |
+
elif relation_label == "EDUCATED_AT":
|
| 117 |
+
subj_type = "PER"
|
| 118 |
+
obj_type = "ORG"
|
| 119 |
+
elif relation_label == "POLITICAL_AFFILIATION":
|
| 120 |
+
subj_type = "PER"
|
| 121 |
+
obj_type = "ORG"
|
| 122 |
+
elif relation_label == "CHILD_OF":
|
| 123 |
+
subj_type = "PER"
|
| 124 |
+
obj_type = "PER"
|
| 125 |
+
elif relation_label == "SPOUSE":
|
| 126 |
+
subj_type = "PER"
|
| 127 |
+
obj_type = "PER"
|
| 128 |
+
elif relation_label == "DATE_FOUNDED":
|
| 129 |
+
subj_type = "ORG"
|
| 130 |
+
obj_type = "DATE"
|
| 131 |
+
elif relation_label == "HEADQUARTERS":
|
| 132 |
+
subj_type = "ORG"
|
| 133 |
+
obj_type = "LOC"
|
| 134 |
+
elif relation_label == "SUBSIDIARY_OF":
|
| 135 |
+
subj_type = "ORG"
|
| 136 |
+
obj_type = "ORG"
|
| 137 |
+
elif relation_label == "FOUNDED_BY":
|
| 138 |
+
subj_type = "ORG"
|
| 139 |
+
obj_type = "PER"
|
| 140 |
+
elif relation_label == "CEO":
|
| 141 |
+
subj_type = "ORG"
|
| 142 |
+
obj_type = "PER"
|
| 143 |
+
else:
|
| 144 |
+
raise ValueError(f"Unknown relation label: {relation_label}")
|
| 145 |
+
return subj_type, obj_type
|
| 146 |
+
|
| 147 |
+
|
| 148 |
+
def remove_contiguous_whitespaces(text):
|
| 149 |
+
# +1 to account for regular whitespace at the beginning
|
| 150 |
+
contiguous_whitespaces_indices = [(m.start(0) + 1, m.end(0)) for m in re.finditer(' +', text)]
|
| 151 |
+
cleaned_text = re.sub(" +", " ", text)
|
| 152 |
+
return cleaned_text, contiguous_whitespaces_indices
|
| 153 |
+
|
| 154 |
+
|
| 155 |
+
def fix_char_index(char_index, contiguous_whitespaces_indices):
|
| 156 |
+
new_char_index = char_index
|
| 157 |
+
offset = 0
|
| 158 |
+
for ws_start, ws_end in contiguous_whitespaces_indices:
|
| 159 |
+
if char_index >= ws_end:
|
| 160 |
+
offset = offset + (ws_end - ws_start)
|
| 161 |
+
new_char_index -= offset
|
| 162 |
+
return new_char_index
|
| 163 |
+
|
| 164 |
+
|
| 165 |
+
class KnowledgeNet(datasets.GeneratorBasedBuilder):
|
| 166 |
+
"""The KnowledgeNet dataset for automatically populating a knowledge base"""
|
| 167 |
+
|
| 168 |
+
BUILDER_CONFIGS = [
|
| 169 |
+
datasets.BuilderConfig(
|
| 170 |
+
name="knet", version=_VERSION, description="The original KnowledgeNet formatted for RE."
|
| 171 |
+
),
|
| 172 |
+
datasets.BuilderConfig(
|
| 173 |
+
name="knet_re", version=_VERSION, description="The original KnowledgeNet formatted for RE."
|
| 174 |
+
),
|
| 175 |
+
datasets.BuilderConfig(
|
| 176 |
+
name="knet_tokenized", version=_VERSION, description="KnowledgeNet tokenized and reformatted."
|
| 177 |
+
),
|
| 178 |
+
]
|
| 179 |
+
|
| 180 |
+
DEFAULT_CONFIG_NAME = "knet" # type: ignore
|
| 181 |
+
|
| 182 |
+
def _info(self):
|
| 183 |
+
if self.config.name == "knet_tokenized":
|
| 184 |
+
features = datasets.Features(
|
| 185 |
+
{
|
| 186 |
+
"doc_id": datasets.Value("string"),
|
| 187 |
+
"passage_id": datasets.Value("string"),
|
| 188 |
+
"fact_id": datasets.Value("string"),
|
| 189 |
+
"tokens": datasets.Sequence(datasets.Value("string")),
|
| 190 |
+
"subj_start": datasets.Value("int32"),
|
| 191 |
+
"subj_end": datasets.Value("int32"),
|
| 192 |
+
"subj_type": datasets.ClassLabel(names=_NER_CLASS_LABELS),
|
| 193 |
+
"subj_uri": datasets.Value("string"),
|
| 194 |
+
"obj_start": datasets.Value("int32"),
|
| 195 |
+
"obj_end": datasets.Value("int32"),
|
| 196 |
+
"obj_type": datasets.ClassLabel(names=_NER_CLASS_LABELS),
|
| 197 |
+
"obj_uri": datasets.Value("string"),
|
| 198 |
+
"relation": datasets.ClassLabel(names=_CLASS_LABELS),
|
| 199 |
+
}
|
| 200 |
+
)
|
| 201 |
+
elif self.config.name == "knet_re":
|
| 202 |
+
features = datasets.Features(
|
| 203 |
+
{
|
| 204 |
+
"documentId": datasets.Value("string"),
|
| 205 |
+
"passageId": datasets.Value("string"),
|
| 206 |
+
"factId": datasets.Value("string"),
|
| 207 |
+
"passageText": datasets.Value("string"),
|
| 208 |
+
"humanReadable": datasets.Value("string"),
|
| 209 |
+
"annotatedPassage": datasets.Value("string"),
|
| 210 |
+
"subjectStart": datasets.Value("int32"),
|
| 211 |
+
"subjectEnd": datasets.Value("int32"),
|
| 212 |
+
"subjectText": datasets.Value("string"),
|
| 213 |
+
"subjectType": datasets.ClassLabel(names=_NER_CLASS_LABELS),
|
| 214 |
+
"subjectUri": datasets.Value("string"),
|
| 215 |
+
"objectStart": datasets.Value("int32"),
|
| 216 |
+
"objectEnd": datasets.Value("int32"),
|
| 217 |
+
"objectText": datasets.Value("string"),
|
| 218 |
+
"objectType": datasets.ClassLabel(names=_NER_CLASS_LABELS),
|
| 219 |
+
"objectUri": datasets.Value("string"),
|
| 220 |
+
"relation": datasets.ClassLabel(names=_CLASS_LABELS),
|
| 221 |
+
}
|
| 222 |
+
)
|
| 223 |
+
else:
|
| 224 |
+
features = datasets.Features(
|
| 225 |
+
{
|
| 226 |
+
"fold": datasets.Value("int32"),
|
| 227 |
+
"documentId": datasets.Value("string"),
|
| 228 |
+
"source": datasets.Value("string"),
|
| 229 |
+
"documentText": datasets.Value("string"),
|
| 230 |
+
"passages": [{
|
| 231 |
+
"passageId": datasets.Value("string"),
|
| 232 |
+
"passageStart": datasets.Value("int32"),
|
| 233 |
+
"passageEnd": datasets.Value("int32"),
|
| 234 |
+
"passageText": datasets.Value("string"),
|
| 235 |
+
"exhaustivelyAnnotatedProperties": [{
|
| 236 |
+
"propertyId": datasets.Value("string"),
|
| 237 |
+
"propertyName": datasets.Value("string"),
|
| 238 |
+
"propertyDescription": datasets.Value("string"),
|
| 239 |
+
}],
|
| 240 |
+
"facts": [{
|
| 241 |
+
"factId": datasets.Value("string"),
|
| 242 |
+
"propertyId": datasets.Value("string"),
|
| 243 |
+
"humanReadable": datasets.Value("string"),
|
| 244 |
+
"annotatedPassage": datasets.Value("string"),
|
| 245 |
+
"subjectStart": datasets.Value("int32"),
|
| 246 |
+
"subjectEnd": datasets.Value("int32"),
|
| 247 |
+
"subjectText": datasets.Value("string"),
|
| 248 |
+
"subjectUri": datasets.Value("string"),
|
| 249 |
+
"objectStart": datasets.Value("int32"),
|
| 250 |
+
"objectEnd": datasets.Value("int32"),
|
| 251 |
+
"objectText": datasets.Value("string"),
|
| 252 |
+
"objectUri": datasets.Value("string"),
|
| 253 |
+
}],
|
| 254 |
+
}],
|
| 255 |
+
}
|
| 256 |
+
)
|
| 257 |
+
|
| 258 |
+
return datasets.DatasetInfo(
|
| 259 |
+
# This is the description that will appear on the datasets page.
|
| 260 |
+
description=_DESCRIPTION,
|
| 261 |
+
# This defines the different columns of the dataset and their types
|
| 262 |
+
features=features, # Here we define them above because they are different between the two configurations
|
| 263 |
+
# If there's a common (input, target) tuple from the features,
|
| 264 |
+
# specify them here. They'll be used if as_supervised=True in
|
| 265 |
+
# builder.as_dataset.
|
| 266 |
+
supervised_keys=None,
|
| 267 |
+
# Homepage of the dataset for documentation
|
| 268 |
+
homepage=_HOMEPAGE,
|
| 269 |
+
# License for the dataset if available
|
| 270 |
+
license=_LICENSE,
|
| 271 |
+
# Citation for the dataset
|
| 272 |
+
citation=_CITATION,
|
| 273 |
+
)
|
| 274 |
+
|
| 275 |
+
def _split_generators(self, dl_manager):
|
| 276 |
+
"""Returns SplitGenerators."""
|
| 277 |
+
# If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
|
| 278 |
+
|
| 279 |
+
# dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLs
|
| 280 |
+
# 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.
|
| 281 |
+
# By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
|
| 282 |
+
|
| 283 |
+
downloaded_files = dl_manager.download_and_extract(_URLS)
|
| 284 |
+
# splits = [datasets.Split.TRAIN, datasets.Split.TEST]
|
| 285 |
+
splits = [datasets.Split.TRAIN]
|
| 286 |
+
return [datasets.SplitGenerator(name=i, gen_kwargs={"filepath": downloaded_files[str(i)], "split": i})
|
| 287 |
+
for i in splits]
|
| 288 |
+
|
| 289 |
+
def _generate_examples(self, filepath, split):
|
| 290 |
+
"""Yields examples."""
|
| 291 |
+
# This method will receive as arguments the `gen_kwargs` defined in the previous `_split_generators` method.
|
| 292 |
+
# It is in charge of opening the given file and yielding (key, example) tuples from the dataset
|
| 293 |
+
# The key is not important, it's more here for legacy reason (legacy from tfds)
|
| 294 |
+
if self.config.name == "knet_tokenized":
|
| 295 |
+
from spacy.lang.en import English
|
| 296 |
+
word_splitter = English()
|
| 297 |
+
else:
|
| 298 |
+
word_splitter = None
|
| 299 |
+
with open(filepath, encoding="utf-8") as f:
|
| 300 |
+
for line in f:
|
| 301 |
+
doc = json.loads(line)
|
| 302 |
+
if self.config.name == "knet":
|
| 303 |
+
yield doc["documentId"], doc
|
| 304 |
+
else:
|
| 305 |
+
for passage in doc["passages"]:
|
| 306 |
+
# Skip passages without facts right away
|
| 307 |
+
if len(passage["facts"]) == 0:
|
| 308 |
+
continue
|
| 309 |
+
|
| 310 |
+
text = passage["passageText"]
|
| 311 |
+
passage_start = passage["passageStart"]
|
| 312 |
+
|
| 313 |
+
if self.config.name == "knet_tokenized":
|
| 314 |
+
cleaned_text, contiguous_ws_indices = remove_contiguous_whitespaces(text)
|
| 315 |
+
spacy_doc = word_splitter(cleaned_text)
|
| 316 |
+
word_tokens = [t.text for t in spacy_doc]
|
| 317 |
+
for fact in passage["facts"]:
|
| 318 |
+
subj_start = fix_char_index(fact["subjectStart"] - passage_start, contiguous_ws_indices)
|
| 319 |
+
subj_end = fix_char_index(fact["subjectEnd"] - passage_start, contiguous_ws_indices)
|
| 320 |
+
obj_start = fix_char_index(fact["objectStart"] - passage_start, contiguous_ws_indices)
|
| 321 |
+
obj_end = fix_char_index(fact["objectEnd"] - passage_start, contiguous_ws_indices)
|
| 322 |
+
# Get exclusive token spans from char spans
|
| 323 |
+
subj_span = spacy_doc.char_span(subj_start, subj_end, alignment_mode="expand")
|
| 324 |
+
obj_span = spacy_doc.char_span(obj_start, obj_end, alignment_mode="expand")
|
| 325 |
+
|
| 326 |
+
relation_label = fact["humanReadable"].split(">")[1][2:]
|
| 327 |
+
subj_type, obj_type = get_entity_types_from_relation(relation_label)
|
| 328 |
+
id_ = fact["factId"]
|
| 329 |
+
|
| 330 |
+
yield id_, {
|
| 331 |
+
"doc_id": doc["documentId"],
|
| 332 |
+
"passage_id": passage["passageId"],
|
| 333 |
+
"fact_id": id_,
|
| 334 |
+
"tokens": word_tokens,
|
| 335 |
+
"subj_start": subj_span.start,
|
| 336 |
+
"subj_end": subj_span.end,
|
| 337 |
+
"subj_type": subj_type,
|
| 338 |
+
"subj_uri": fact["subjectUri"],
|
| 339 |
+
"obj_start": obj_span.start,
|
| 340 |
+
"obj_end": obj_span.end,
|
| 341 |
+
"obj_type": obj_type,
|
| 342 |
+
"obj_uri": fact["objectUri"],
|
| 343 |
+
"relation": relation_label
|
| 344 |
+
}
|
| 345 |
+
else:
|
| 346 |
+
for fact in passage["facts"]:
|
| 347 |
+
relation_label = fact["humanReadable"].split(">")[1][2:]
|
| 348 |
+
subj_type, obj_type = get_entity_types_from_relation(relation_label)
|
| 349 |
+
id_ = fact["factId"]
|
| 350 |
+
yield id_, {
|
| 351 |
+
"documentId": doc["documentId"],
|
| 352 |
+
"passageId": passage["passageId"],
|
| 353 |
+
"passageText": passage["passageText"],
|
| 354 |
+
"factId": id_,
|
| 355 |
+
"humanReadable": fact["humanReadable"],
|
| 356 |
+
"annotatedPassage": fact["annotatedPassage"],
|
| 357 |
+
"subjectStart": fact["subjectStart"] - passage_start,
|
| 358 |
+
"subjectEnd": fact["subjectEnd"] - passage_start,
|
| 359 |
+
"subjectText": fact["subjectText"],
|
| 360 |
+
"subjectType": subj_type,
|
| 361 |
+
"subjectUri": fact["subjectUri"],
|
| 362 |
+
"objectStart": fact["objectStart"] - passage_start,
|
| 363 |
+
"objectEnd": fact["objectEnd"] - passage_start,
|
| 364 |
+
"objectText": fact["objectText"],
|
| 365 |
+
"objectType": obj_type,
|
| 366 |
+
"objectUri": fact["objectUri"],
|
| 367 |
+
"relation": relation_label
|
| 368 |
+
}
|