metadata
annotations_creators:
- found
language_creators:
- found
language:
- ca
- de
- el
- en
- eo
- es
- fi
- fr
- hu
- it
- nl
- 'no'
- pl
- pt
- ru
- sv
license:
- unknown
multilinguality:
- multilingual
size_categories:
- 1K<n<10K
source_datasets:
- original
task_categories:
- translation
task_ids: []
paperswithcode_id: null
pretty_name: OpusBooks
dataset_info:
- config_name: ca-de
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- ca
- de
splits:
- name: train
num_bytes: 899565
num_examples: 4445
download_size: 349126
dataset_size: 899565
- config_name: ca-en
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- ca
- en
splits:
- name: train
num_bytes: 863174
num_examples: 4605
download_size: 336276
dataset_size: 863174
- config_name: de-en
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- en
splits:
- name: train
num_bytes: 13739047
num_examples: 51467
download_size: 5124458
dataset_size: 13739047
- config_name: el-en
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- el
- en
splits:
- name: train
num_bytes: 552579
num_examples: 1285
download_size: 175537
dataset_size: 552579
- config_name: de-eo
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- eo
splits:
- name: train
num_bytes: 398885
num_examples: 1363
download_size: 150822
dataset_size: 398885
- config_name: en-eo
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- eo
splits:
- name: train
num_bytes: 386231
num_examples: 1562
download_size: 145339
dataset_size: 386231
- config_name: de-es
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- es
splits:
- name: train
num_bytes: 7592487
num_examples: 27526
download_size: 2802010
dataset_size: 7592487
- config_name: el-es
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- el
- es
splits:
- name: train
num_bytes: 527991
num_examples: 1096
download_size: 168306
dataset_size: 527991
- config_name: en-es
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- es
splits:
- name: train
num_bytes: 25291783
num_examples: 93470
download_size: 9257150
dataset_size: 25291783
- config_name: eo-es
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- eo
- es
splits:
- name: train
num_bytes: 409591
num_examples: 1677
download_size: 154950
dataset_size: 409591
- config_name: en-fi
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- fi
splits:
- name: train
num_bytes: 715039
num_examples: 3645
download_size: 266714
dataset_size: 715039
- config_name: es-fi
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- es
- fi
splits:
- name: train
num_bytes: 710462
num_examples: 3344
download_size: 264316
dataset_size: 710462
- config_name: de-fr
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- fr
splits:
- name: train
num_bytes: 9544399
num_examples: 34916
download_size: 3556168
dataset_size: 9544399
- config_name: el-fr
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- el
- fr
splits:
- name: train
num_bytes: 539933
num_examples: 1237
download_size: 169241
dataset_size: 539933
- config_name: en-fr
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- fr
splits:
- name: train
num_bytes: 32997199
num_examples: 127085
download_size: 12009501
dataset_size: 32997199
- config_name: eo-fr
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- eo
- fr
splits:
- name: train
num_bytes: 412999
num_examples: 1588
download_size: 152040
dataset_size: 412999
- config_name: es-fr
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- es
- fr
splits:
- name: train
num_bytes: 14382198
num_examples: 56319
download_size: 5203099
dataset_size: 14382198
- config_name: fi-fr
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fi
- fr
splits:
- name: train
num_bytes: 746097
num_examples: 3537
download_size: 276633
dataset_size: 746097
- config_name: ca-hu
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- ca
- hu
splits:
- name: train
num_bytes: 886162
num_examples: 4463
download_size: 346425
dataset_size: 886162
- config_name: de-hu
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- hu
splits:
- name: train
num_bytes: 13515043
num_examples: 51780
download_size: 5069455
dataset_size: 13515043
- config_name: el-hu
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- el
- hu
splits:
- name: train
num_bytes: 546290
num_examples: 1090
download_size: 176715
dataset_size: 546290
- config_name: en-hu
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- hu
splits:
- name: train
num_bytes: 35256934
num_examples: 137151
download_size: 13232578
dataset_size: 35256934
- config_name: eo-hu
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- eo
- hu
splits:
- name: train
num_bytes: 389112
num_examples: 1636
download_size: 151332
dataset_size: 389112
- config_name: fr-hu
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- hu
splits:
- name: train
num_bytes: 22483133
num_examples: 89337
download_size: 8328639
dataset_size: 22483133
- config_name: de-it
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- it
splits:
- name: train
num_bytes: 7760020
num_examples: 27381
download_size: 2811066
dataset_size: 7760020
- config_name: en-it
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- it
splits:
- name: train
num_bytes: 8993803
num_examples: 32332
download_size: 3295251
dataset_size: 8993803
- config_name: eo-it
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- eo
- it
splits:
- name: train
num_bytes: 387606
num_examples: 1453
download_size: 146899
dataset_size: 387606
- config_name: es-it
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- es
- it
splits:
- name: train
num_bytes: 7837703
num_examples: 28868
download_size: 2864028
dataset_size: 7837703
- config_name: fr-it
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- it
splits:
- name: train
num_bytes: 4752171
num_examples: 14692
download_size: 1737670
dataset_size: 4752171
- config_name: hu-it
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- hu
- it
splits:
- name: train
num_bytes: 8445585
num_examples: 30949
download_size: 3101681
dataset_size: 8445585
- config_name: ca-nl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- ca
- nl
splits:
- name: train
num_bytes: 884823
num_examples: 4329
download_size: 340308
dataset_size: 884823
- config_name: de-nl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- nl
splits:
- name: train
num_bytes: 3561764
num_examples: 15622
download_size: 1325189
dataset_size: 3561764
- config_name: en-nl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- nl
splits:
- name: train
num_bytes: 10278038
num_examples: 38652
download_size: 3727995
dataset_size: 10278038
- config_name: es-nl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- es
- nl
splits:
- name: train
num_bytes: 9062389
num_examples: 32247
download_size: 3245558
dataset_size: 9062389
- config_name: fr-nl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- nl
splits:
- name: train
num_bytes: 10408148
num_examples: 40017
download_size: 3720151
dataset_size: 10408148
- config_name: hu-nl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- hu
- nl
splits:
- name: train
num_bytes: 10814173
num_examples: 43428
download_size: 3998988
dataset_size: 10814173
- config_name: it-nl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- it
- nl
splits:
- name: train
num_bytes: 1328305
num_examples: 2359
download_size: 476875
dataset_size: 1328305
- config_name: en-no
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- 'no'
splits:
- name: train
num_bytes: 661978
num_examples: 3499
download_size: 246977
dataset_size: 661978
- config_name: es-no
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- es
- 'no'
splits:
- name: train
num_bytes: 729125
num_examples: 3585
download_size: 270796
dataset_size: 729125
- config_name: fi-no
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fi
- 'no'
splits:
- name: train
num_bytes: 691181
num_examples: 3414
download_size: 256267
dataset_size: 691181
- config_name: fr-no
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- 'no'
splits:
- name: train
num_bytes: 692786
num_examples: 3449
download_size: 256501
dataset_size: 692786
- config_name: hu-no
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- hu
- 'no'
splits:
- name: train
num_bytes: 695497
num_examples: 3410
download_size: 267047
dataset_size: 695497
- config_name: en-pl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- pl
splits:
- name: train
num_bytes: 583091
num_examples: 2831
download_size: 226855
dataset_size: 583091
- config_name: fi-pl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fi
- pl
splits:
- name: train
num_bytes: 613791
num_examples: 2814
download_size: 236123
dataset_size: 613791
- config_name: fr-pl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- pl
splits:
- name: train
num_bytes: 614248
num_examples: 2825
download_size: 235905
dataset_size: 614248
- config_name: hu-pl
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- hu
- pl
splits:
- name: train
num_bytes: 616161
num_examples: 2859
download_size: 245670
dataset_size: 616161
- config_name: de-pt
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- pt
splits:
- name: train
num_bytes: 317155
num_examples: 1102
download_size: 116319
dataset_size: 317155
- config_name: en-pt
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- pt
splits:
- name: train
num_bytes: 309689
num_examples: 1404
download_size: 111837
dataset_size: 309689
- config_name: eo-pt
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- eo
- pt
splits:
- name: train
num_bytes: 311079
num_examples: 1259
download_size: 116157
dataset_size: 311079
- config_name: es-pt
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- es
- pt
splits:
- name: train
num_bytes: 326884
num_examples: 1327
download_size: 120549
dataset_size: 326884
- config_name: fr-pt
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- pt
splits:
- name: train
num_bytes: 324616
num_examples: 1263
download_size: 115920
dataset_size: 324616
- config_name: hu-pt
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- hu
- pt
splits:
- name: train
num_bytes: 302972
num_examples: 1184
download_size: 115002
dataset_size: 302972
- config_name: it-pt
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- it
- pt
splits:
- name: train
num_bytes: 301428
num_examples: 1163
download_size: 111050
dataset_size: 301428
- config_name: de-ru
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- de
- ru
splits:
- name: train
num_bytes: 5764673
num_examples: 17373
download_size: 1799371
dataset_size: 5764673
- config_name: en-ru
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- ru
splits:
- name: train
num_bytes: 5190880
num_examples: 17496
download_size: 1613419
dataset_size: 5190880
- config_name: es-ru
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- es
- ru
splits:
- name: train
num_bytes: 5281130
num_examples: 16793
download_size: 1648606
dataset_size: 5281130
- config_name: fr-ru
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- ru
splits:
- name: train
num_bytes: 2474210
num_examples: 8197
download_size: 790541
dataset_size: 2474210
- config_name: hu-ru
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- hu
- ru
splits:
- name: train
num_bytes: 7818688
num_examples: 26127
download_size: 2469765
dataset_size: 7818688
- config_name: it-ru
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- it
- ru
splits:
- name: train
num_bytes: 5316952
num_examples: 17906
download_size: 1620478
dataset_size: 5316952
- config_name: en-sv
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- en
- sv
splits:
- name: train
num_bytes: 790785
num_examples: 3095
download_size: 304975
dataset_size: 790785
- config_name: fr-sv
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- fr
- sv
splits:
- name: train
num_bytes: 833553
num_examples: 3002
download_size: 321660
dataset_size: 833553
- config_name: it-sv
features:
- name: id
dtype: string
- name: translation
dtype:
translation:
languages:
- it
- sv
splits:
- name: train
num_bytes: 811413
num_examples: 2998
download_size: 307821
dataset_size: 811413
Dataset Card for OpusBooks
Table of Contents
- Dataset Description
- Dataset Structure
- Dataset Creation
- Considerations for Using the Data
- Additional Information
Dataset Description
- Homepage: http://opus.nlpl.eu/Books.php
- Repository: None
- Paper: http://www.lrec-conf.org/proceedings/lrec2012/pdf/463_Paper.pdf
- Leaderboard: [More Information Needed]
- Point of Contact: [More Information Needed]
Dataset Summary
[More Information Needed]
Supported Tasks and Leaderboards
[More Information Needed]
Languages
[More Information Needed]
Dataset Structure
Data Instances
Here are some examples of questions and facts:
Data Fields
[More Information Needed]
Data Splits
[More Information Needed]
Dataset Creation
Curation Rationale
[More Information Needed]
Source Data
[More Information Needed]
Initial Data Collection and Normalization
[More Information Needed]
Who are the source language producers?
[More Information Needed]
Annotations
[More Information Needed]
Annotation process
[More Information Needed]
Who are the annotators?
[More Information Needed]
Personal and Sensitive Information
[More Information Needed]
Considerations for Using the Data
Social Impact of Dataset
[More Information Needed]
Discussion of Biases
[More Information Needed]
Other Known Limitations
[More Information Needed]
Additional Information
Dataset Curators
[More Information Needed]
Licensing Information
[More Information Needed]
Citation Information
[More Information Needed]
Contributions
Thanks to @abhishekkrthakur for adding this dataset.