Datasets:

Modalities:
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
Dask
License:
Jeronymous commited on
Commit
3616b36
·
verified ·
1 Parent(s): e70ffb7

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +69 -69
README.md CHANGED
@@ -525,13 +525,13 @@ Examples of metadata (except from `text`) are shown for each source in [metadata
525
 
526
  ### Dataset Composition
527
 
528
- The following figure show the distribution of the dataset by language (hatch patterns) and document category (colors).
529
 
530
  ![Dataset composition](figures/fig_dataset_composition.png)
531
 
532
  The following table provides an overview of the dataset composition,
533
- broken down by source and language,
534
- sources being grouped by category.
535
  The table provides the number of documents, words, tokens, and characters for each subset.
536
 
537
  <!-- The following is automatically generated. Do not update manually. -->
@@ -565,7 +565,7 @@ The table provides the number of documents, words, tokens, and characters for ea
565
  <td>583.687</td>
566
  <td>928.618</td>
567
  <td>3619.672</td>
568
- <td><strong>RedPajama</strong> (79.8 %), <strong>GallicaPress</strong> (13.1 %), <strong>GallicaMonographies</strong> (2.71 %), <strong>HAL</strong> (1.75 %), <strong>Theses</strong> (1.51 %), <strong>OpenEdition</strong> (0.388 %), <strong>Wikipedia</strong> (0.317 %), <strong>OpenData</strong> (0.130 %), <strong>wikisource</strong> (0.0856 %), <strong>Gutenberg</strong> (0.0412 %), <strong>YouTube</strong> (0.0362 %), <strong>Claire</strong> (0.0335 %), <strong>DiscoursPublics</strong> (0.0256 %), <strong>InterventionsParlement</strong> (0.0169 %), <strong>QuestionsEcritesParlement</strong> (0.0168 %), <strong>LEGI</strong> (0.0156 %), <strong>wiktionary</strong> (0.0126 %), <strong>AmendementsParlement</strong> (0.00795 %), <strong>Europarl</strong> (0.00772 %)</td>
569
  </tr>
570
  <tr>
571
 
@@ -574,16 +574,16 @@ The table provides the number of documents, words, tokens, and characters for ea
574
  <td>412.202</td>
575
  <td>611.894</td>
576
  <td>2553.541</td>
577
- <td><strong>FineWebEdu</strong> (76.5 %), <strong>PeS2o</strong> (10.7 %), <strong>AmericanStories</strong> (2.34 %), <strong>Pile (FreeLaw)</strong> (2.29 %), <strong>Pile (StackExchange)</strong> (1.68 %), <strong>MathPile</strong> (1.57 %), <strong>Wikipedia</strong> (1.29 %), <strong>Gutenberg</strong> (0.901 %), <strong>Pile (USPTO_Backgrounds)</strong> (0.834 %), <strong>Pile (DM_Mathematics)</strong> (0.805 %), <strong>Eurovoc</strong> (0.420 %), <strong>Pile (Ubuntu_IRC)</strong> (0.353 %), <strong>Claire</strong> (0.190 %), <strong>Pile (PhilPapers)</strong> (0.101 %), <strong>Pile (NIH_ExPorter)</strong> (0.0704 %), <strong>Europarl</strong> (0.0113 %), <strong>Stac</strong> (0.0000198 %)</td>
578
  </tr>
579
  <tr>
580
 
581
- <td><strong>Programming Languages (code)</strong></td>
582
  <td>125.769</td>
583
  <td>51.306</td>
584
  <td>228.954</td>
585
  <td>630.749</td>
586
- <td><strong>JAVASCRIPT</strong> (25.6 %), <strong>JAVA</strong> (12.1 %), <strong>C</strong> (10.5 %), <strong>PHP</strong> (10.0 %), <strong>PYTHON</strong> (9.47 %), <strong>C++</strong> (8.23 %), <strong>C#</strong> (5.84 %), <strong>GO</strong> (4.48 %), <strong>TYPESCRIPT</strong> (4.30 %), <strong>RUST</strong> (1.42 %), <strong>RUBY</strong> (1.04 %), <strong>SWIFT</strong> (0.819 %), <strong>KOTLIN</strong> (0.768 %), <strong>SCALA</strong> (0.693 %), <strong>TEX</strong> (0.658 %), <strong>LUA</strong> (0.597 %), <strong>DART</strong> (0.542 %), <strong>PERL</strong> (0.502 %), <strong>MATHEMATICA</strong> (0.488 %), <strong>ASSEMBLY</strong> (0.379 %), <strong>HASKELL</strong> (0.352 %), <strong>FORTRAN</strong> (0.341 %), <strong>JULIA</strong> (0.288 %), <strong>OCAML</strong> (0.188 %), <strong>ERLANG</strong> (0.114 %), <strong>ELIXIR</strong> (0.113 %), <strong>CLOJURE</strong> (0.0782 %), <strong>R</strong> (0.0690 %), <strong>MATLAB</strong> (0.0187 %), <strong>RACKET</strong> (0.00668 %)</td>
587
  </tr>
588
  <tr>
589
 
@@ -592,7 +592,7 @@ The table provides the number of documents, words, tokens, and characters for ea
592
  <td>105.609</td>
593
  <td>206.610</td>
594
  <td>764.779</td>
595
- <td><strong>RedPajama</strong> (97.5 %), <strong>Wikipedia</strong> (1.68 %), <strong>Eurovoc</strong> (0.725 %), <strong>Gutenberg</strong> (0.0934 %), <strong>Europarl</strong> (0.0355 %)</td>
596
  </tr>
597
  <tr>
598
 
@@ -601,7 +601,7 @@ The table provides the number of documents, words, tokens, and characters for ea
601
  <td>123.857</td>
602
  <td>200.825</td>
603
  <td>759.457</td>
604
- <td><strong>RedPajama</strong> (98.2 %), <strong>Wikipedia</strong> (1.06 %), <strong>Eurovoc</strong> (0.703 %), <strong>Gutenberg</strong> (0.0458 %), <strong>Europarl</strong> (0.0365 %)</td>
605
  </tr>
606
  <tr>
607
 
@@ -610,7 +610,7 @@ The table provides the number of documents, words, tokens, and characters for ea
610
  <td>62.051</td>
611
  <td>112.031</td>
612
  <td>404.454</td>
613
- <td><strong>RedPajama</strong> (96.8 %), <strong>Wikipedia</strong> (1.75 %), <strong>Eurovoc</strong> (1.36 %), <strong>Gutenberg</strong> (0.115 %)</td>
614
  </tr>
615
  <tr>
616
 
@@ -619,7 +619,7 @@ The table provides the number of documents, words, tokens, and characters for ea
619
  <td>17.016</td>
620
  <td>25.494</td>
621
  <td>107.658</td>
622
- <td><strong>CroissantAligned</strong> (99.4 %), <strong>EuroparlAligned</strong> (0.561 %)</td>
623
  </tr>
624
  <tr>
625
 
@@ -628,7 +628,7 @@ The table provides the number of documents, words, tokens, and characters for ea
628
  <td>0.100</td>
629
  <td>0.151</td>
630
  <td>0.638</td>
631
- <td><strong>EuroparlAligned</strong></td>
632
  </tr>
633
  <tr>
634
 
@@ -637,7 +637,7 @@ The table provides the number of documents, words, tokens, and characters for ea
637
  <td>0.103</td>
638
  <td>0.143</td>
639
  <td>0.631</td>
640
- <td><strong>EuroparlAligned</strong></td>
641
  </tr>
642
  <tr>
643
 
@@ -646,7 +646,7 @@ The table provides the number of documents, words, tokens, and characters for ea
646
  <td>0.0908</td>
647
  <td>0.141</td>
648
  <td>0.621</td>
649
- <td><strong>EuroparlAligned</strong></td>
650
  </tr>
651
  <tr>
652
  <td colspan="7"><h4 id="category-web">Category: Web</h4></td></tr>
@@ -657,7 +657,7 @@ The table provides the number of documents, words, tokens, and characters for ea
657
  <td>477.758</td>
658
  <td>741.023</td>
659
  <td>2974.596</td>
660
- <td><strong>2023</strong> (5.63 %), <strong>2022</strong> (13.4 %), <strong>2021</strong> (17.2 %), <strong>2020</strong> (15.8 %), <strong>2019</strong> (18.2 %), <strong>2018</strong> (17.1 %), <strong>2017</strong> (11.7 %), <strong>2016</strong> (0.437 %), <strong>2015</strong> (0.167 %), <strong>2014</strong> (0.284 %)</td>
661
  </tr>
662
  <tr>
663
 
@@ -666,7 +666,7 @@ The table provides the number of documents, words, tokens, and characters for ea
666
  <td>103.078</td>
667
  <td>201.371</td>
668
  <td>747.631</td>
669
- <td><strong>2023</strong> (23.9 %), <strong>2022</strong> (59.0 %), <strong>2021</strong> (17.1 %)</td>
670
  </tr>
671
  <tr>
672
 
@@ -675,7 +675,7 @@ The table provides the number of documents, words, tokens, and characters for ea
675
  <td>121.751</td>
676
  <td>197.125</td>
677
  <td>746.984</td>
678
- <td><strong>2023</strong> (23.7 %), <strong>2022</strong> (59.2 %), <strong>2021</strong> (17.1 %)</td>
679
  </tr>
680
  <tr>
681
 
@@ -684,7 +684,7 @@ The table provides the number of documents, words, tokens, and characters for ea
684
  <td>60.194</td>
685
  <td>108.416</td>
686
  <td>393.012</td>
687
- <td><strong>2023</strong> (23.9 %), <strong>2022</strong> (59.0 %), <strong>2021</strong> (17.1 %)</td>
688
  </tr>
689
  <tr>
690
  <td><a href="#finewebedu"><strong>FineWebEdu</strong></a></td>
@@ -693,7 +693,7 @@ The table provides the number of documents, words, tokens, and characters for ea
693
  <td>327.453</td>
694
  <td>467.837</td>
695
  <td>2018.215</td>
696
- <td><strong>2024</strong> (2.73 %), <strong>2023</strong> (18.9 %), <strong>2022</strong> (18.0 %), <strong>2021</strong> (22.3 %), <strong>2020</strong> (18.1 %), <strong>2019</strong> (20.0 %)</td>
697
  </tr>
698
  <tr>
699
  <td colspan="7"><h4 id="category-newspaper">Category: Newspaper</h4></td></tr>
@@ -713,7 +713,7 @@ The table provides the number of documents, words, tokens, and characters for ea
713
  <td>8.902</td>
714
  <td>14.313</td>
715
  <td>50.844</td>
716
- <td></td>
717
  </tr>
718
  <tr>
719
  <td colspan="7"><h4 id="category-technical">Category: Technical</h4></td></tr>
@@ -837,53 +837,6 @@ The table provides the number of documents, words, tokens, and characters for ea
837
  <td></td>
838
  </tr>
839
  <tr>
840
- <td colspan="7"><h4 id="category-multilingual-parallel-corpora">Category: Multilingual Parallel Corpora</h4></td></tr>
841
- <tr>
842
- <td><a href="#croissantaligned"><strong>CroissantAligned</strong></a></td>
843
- <td><strong>fr-en</strong></td>
844
- <td>408.029</td>
845
- <td>16.911</td>
846
- <td>25.351</td>
847
- <td>107.003</td>
848
- <td></td>
849
- </tr>
850
- <tr>
851
- <td rowspan="4" style="vertical-align: top;"><a href="#europarl-monolingual-and-parallel"><strong>EuroparlAligned</strong></a></td>
852
- <td><strong>it-en</strong></td>
853
- <td>1.901</td>
854
- <td>0.100</td>
855
- <td>0.151</td>
856
- <td>0.638</td>
857
- <td></td>
858
- </tr>
859
- <tr>
860
-
861
- <td><strong>fr-en</strong></td>
862
- <td>2.003</td>
863
- <td>0.105</td>
864
- <td>0.143</td>
865
- <td>0.655</td>
866
- <td></td>
867
- </tr>
868
- <tr>
869
-
870
- <td><strong>es-en</strong></td>
871
- <td>1.961</td>
872
- <td>0.103</td>
873
- <td>0.143</td>
874
- <td>0.631</td>
875
- <td></td>
876
- </tr>
877
- <tr>
878
-
879
- <td><strong>de-fr</strong></td>
880
- <td>1.792</td>
881
- <td>0.0908</td>
882
- <td>0.141</td>
883
- <td>0.621</td>
884
- <td></td>
885
- </tr>
886
- <tr>
887
  <td colspan="7"><h4 id="category-legislative-texts">Category: Legislative Texts</h4></td></tr>
888
  <tr>
889
  <td><a href="#pile-uncopyrighted"><strong>Pile (FreeLaw)</strong></a></td>
@@ -1136,7 +1089,7 @@ The table provides the number of documents, words, tokens, and characters for ea
1136
  <td>0.818</td>
1137
  <td>1.161</td>
1138
  <td>4.709</td>
1139
- <td><strong>MediaSum</strong> (87.2 %), <strong>DialogStudio</strong> (10.2 %), <strong>BNC</strong> (1.46 %), <strong>OANC</strong> (0.690 %), <strong>DailyDialog</strong> (0.199 %), <strong>ICSI</strong> (0.156 %), <strong>AMI</strong> (0.144 %)</td>
1140
  </tr>
1141
  <tr>
1142
 
@@ -1145,7 +1098,7 @@ The table provides the number of documents, words, tokens, and characters for ea
1145
  <td>0.210</td>
1146
  <td>0.311</td>
1147
  <td>1.314</td>
1148
- <td><strong>AssembleeNationale</strong> (60.1 %), <strong>Senat</strong> (24.5 %), <strong>Theatre</strong> (9.33 %), <strong>ESLO</strong> (2.84 %), <strong>ORFEO</strong> (0.704 %), <strong>SUMM</strong> (0.601 %), <strong>TCOF</strong> (0.421 %), <strong>CFPP</strong> (0.367 %), <strong>OFROM</strong> (0.320 %), <strong>PFC</strong> (0.296 %), <strong>FREDSum</strong> (0.186 %), <strong>CLAPI</strong> (0.0974 %), <strong>CID</strong> (0.0756 %), <strong>LINAGORA</strong> (0.0605 %), <strong>ACSYNT</strong> (0.0309 %), <strong>OTG</strong> (0.0200 %), <strong>Rhapsodie</strong> (0.0152 %), <strong>ParisStories</strong> (0.0134 %), <strong>UBS</strong> (0.00524 %)</td>
1149
  </tr>
1150
  <tr>
1151
  <td><a href="#youtube"><strong>YouTube</strong></a></td>
@@ -1166,6 +1119,53 @@ The table provides the number of documents, words, tokens, and characters for ea
1166
  <td></td>
1167
  </tr>
1168
  <tr>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1169
  <td colspan="7"><h4 id="category-programming">Category: Programming</h4></td></tr>
1170
  <tr>
1171
  <td rowspan="30" style="vertical-align: top;"><a href="#thestack"><strong>TheStack</strong></a></td>
 
525
 
526
  ### Dataset Composition
527
 
528
+ The following figure shows the distribution of the dataset by language (colors) and category (hatch patterns).
529
 
530
  ![Dataset composition](figures/fig_dataset_composition.png)
531
 
532
  The following table provides an overview of the dataset composition,
533
+ broken down by source and language.
534
+ Sources are grouped by category.
535
  The table provides the number of documents, words, tokens, and characters for each subset.
536
 
537
  <!-- The following is automatically generated. Do not update manually. -->
 
565
  <td>583.687</td>
566
  <td>928.618</td>
567
  <td>3619.672</td>
568
+ <td><a href="figures/fig_distribution_french_pie.png">composition details</a></td>
569
  </tr>
570
  <tr>
571
 
 
574
  <td>412.202</td>
575
  <td>611.894</td>
576
  <td>2553.541</td>
577
+ <td><a href="figures/fig_distribution_english_pie.png">composition details</a></td>
578
  </tr>
579
  <tr>
580
 
581
+ <td><strong>code</strong></td>
582
  <td>125.769</td>
583
  <td>51.306</td>
584
  <td>228.954</td>
585
  <td>630.749</td>
586
+ <td><a href="figures/fig_distribution_code_pie.png">composition details</a></td>
587
  </tr>
588
  <tr>
589
 
 
592
  <td>105.609</td>
593
  <td>206.610</td>
594
  <td>764.779</td>
595
+ <td><a href="figures/fig_distribution_german_pie.png">composition details</a></td>
596
  </tr>
597
  <tr>
598
 
 
601
  <td>123.857</td>
602
  <td>200.825</td>
603
  <td>759.457</td>
604
+ <td><a href="figures/fig_distribution_spanish_pie.png">composition details</a></td>
605
  </tr>
606
  <tr>
607
 
 
610
  <td>62.051</td>
611
  <td>112.031</td>
612
  <td>404.454</td>
613
+ <td><a href="figures/fig_distribution_italian_pie.png">composition details</a></td>
614
  </tr>
615
  <tr>
616
 
 
619
  <td>17.016</td>
620
  <td>25.494</td>
621
  <td>107.658</td>
622
+ <td><a href="figures/fig_distribution_fr-en_pie.png">composition details</a></td>
623
  </tr>
624
  <tr>
625
 
 
628
  <td>0.100</td>
629
  <td>0.151</td>
630
  <td>0.638</td>
631
+ <td></td>
632
  </tr>
633
  <tr>
634
 
 
637
  <td>0.103</td>
638
  <td>0.143</td>
639
  <td>0.631</td>
640
+ <td></td>
641
  </tr>
642
  <tr>
643
 
 
646
  <td>0.0908</td>
647
  <td>0.141</td>
648
  <td>0.621</td>
649
+ <td></td>
650
  </tr>
651
  <tr>
652
  <td colspan="7"><h4 id="category-web">Category: Web</h4></td></tr>
 
657
  <td>477.758</td>
658
  <td>741.023</td>
659
  <td>2974.596</td>
660
+ <td><a href="figures/fig_distribution_redpajama-french_histogram.png">composition details</a></td>
661
  </tr>
662
  <tr>
663
 
 
666
  <td>103.078</td>
667
  <td>201.371</td>
668
  <td>747.631</td>
669
+ <td><a href="figures/fig_distribution_redpajama-german_histogram.png">composition details</a></td>
670
  </tr>
671
  <tr>
672
 
 
675
  <td>121.751</td>
676
  <td>197.125</td>
677
  <td>746.984</td>
678
+ <td><a href="figures/fig_distribution_redpajama-spanish_histogram.png">composition details</a></td>
679
  </tr>
680
  <tr>
681
 
 
684
  <td>60.194</td>
685
  <td>108.416</td>
686
  <td>393.012</td>
687
+ <td><a href="figures/fig_distribution_redpajama-italian_histogram.png">composition details</a></td>
688
  </tr>
689
  <tr>
690
  <td><a href="#finewebedu"><strong>FineWebEdu</strong></a></td>
 
693
  <td>327.453</td>
694
  <td>467.837</td>
695
  <td>2018.215</td>
696
+ <td><a href="figures/fig_distribution_finewebedu-english_histogram.png">composition details</a></td>
697
  </tr>
698
  <tr>
699
  <td colspan="7"><h4 id="category-newspaper">Category: Newspaper</h4></td></tr>
 
713
  <td>8.902</td>
714
  <td>14.313</td>
715
  <td>50.844</td>
716
+ <td><a href="figures/fig_distribution_americanstories_histogram.png">composition details</a></td>
717
  </tr>
718
  <tr>
719
  <td colspan="7"><h4 id="category-technical">Category: Technical</h4></td></tr>
 
837
  <td></td>
838
  </tr>
839
  <tr>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
840
  <td colspan="7"><h4 id="category-legislative-texts">Category: Legislative Texts</h4></td></tr>
841
  <tr>
842
  <td><a href="#pile-uncopyrighted"><strong>Pile (FreeLaw)</strong></a></td>
 
1089
  <td>0.818</td>
1090
  <td>1.161</td>
1091
  <td>4.709</td>
1092
+ <td><a href="figures/fig_distribution_claire-english_pie.png">composition details</a></td>
1093
  </tr>
1094
  <tr>
1095
 
 
1098
  <td>0.210</td>
1099
  <td>0.311</td>
1100
  <td>1.314</td>
1101
+ <td><a href="figures/fig_distribution_claire-french_pie.png">composition details</a></td>
1102
  </tr>
1103
  <tr>
1104
  <td><a href="#youtube"><strong>YouTube</strong></a></td>
 
1119
  <td></td>
1120
  </tr>
1121
  <tr>
1122
+ <td colspan="7"><h4 id="category-multilingual-parallel-corpora">Category: Multilingual Parallel Corpora</h4></td></tr>
1123
+ <tr>
1124
+ <td><a href="#croissantaligned"><strong>CroissantAligned</strong></a></td>
1125
+ <td><strong>fr-en</strong></td>
1126
+ <td>408.029</td>
1127
+ <td>16.911</td>
1128
+ <td>25.351</td>
1129
+ <td>107.003</td>
1130
+ <td></td>
1131
+ </tr>
1132
+ <tr>
1133
+ <td rowspan="4" style="vertical-align: top;"><a href="#europarl-monolingual-and-parallel"><strong>EuroparlAligned</strong></a></td>
1134
+ <td><strong>it-en</strong></td>
1135
+ <td>1.901</td>
1136
+ <td>0.100</td>
1137
+ <td>0.151</td>
1138
+ <td>0.638</td>
1139
+ <td></td>
1140
+ </tr>
1141
+ <tr>
1142
+
1143
+ <td><strong>fr-en</strong></td>
1144
+ <td>2.003</td>
1145
+ <td>0.105</td>
1146
+ <td>0.143</td>
1147
+ <td>0.655</td>
1148
+ <td></td>
1149
+ </tr>
1150
+ <tr>
1151
+
1152
+ <td><strong>es-en</strong></td>
1153
+ <td>1.961</td>
1154
+ <td>0.103</td>
1155
+ <td>0.143</td>
1156
+ <td>0.631</td>
1157
+ <td></td>
1158
+ </tr>
1159
+ <tr>
1160
+
1161
+ <td><strong>de-fr</strong></td>
1162
+ <td>1.792</td>
1163
+ <td>0.0908</td>
1164
+ <td>0.141</td>
1165
+ <td>0.621</td>
1166
+ <td></td>
1167
+ </tr>
1168
+ <tr>
1169
  <td colspan="7"><h4 id="category-programming">Category: Programming</h4></td></tr>
1170
  <tr>
1171
  <td rowspan="30" style="vertical-align: top;"><a href="#thestack"><strong>TheStack</strong></a></td>