Hugging Face's logo

Datasets:
credi-net
/
DomainPool

License:
Dataset card Data Studio Files
xet
 Community
ekmpa commited on
Commit
be1aedc
·
verified ·
1 Parent(s): 4cebf28

added plots

Browse files
Files changed (1) hide show
  1. README.md +234 -323
README.md CHANGED
@@ -2,36 +2,27 @@
2
  license: cc-by-4.0
3
  ---
4
 
5
- > **( Important! )**
6
- > This dataset card is currently under construction and is incomplete.
7
-
8
-
9
  # Dataset Card for *`domain-pool`*
10
 
11
 
12
  `domain-pool` is a fine grained and cross-domain aggregate labelled set of web domains. Its default form has majority categories downsampled to present a more balanced set of
13
  148,830 domains, while the full imbalanced set has 5,671,355 datapoints.
14
  These web domains are mainly labelled across three axes:
15
- - reliability labels form the principal category. Reliability scores can be numerical, then normalized to [0.0,1.0] where higher is better; or categorical. Categories are listed below.
16
- - factuality labels are sparser, spanning X. They are categorical: a domain can have low, medium or high factuality, as assessed by sources like fact-checking organisations.
17
  - bias labels can, like the reliability ones, either be a continuous score on [0.0,1.0], or a category (can either be quantitative or across left/right axis).
18
  All domains also have the original data source indicated, and the dataset's domain scope (e.g., misinformation, or malware; domain-pool spans Y domains).
19
  A large part of these data sources are open-sources academic datasets, and the labels for Y domains were also collected manually
20
  from online sources (governmental, journalistic or academic) that gathered domain lists in non-machine readable format.
21
 
22
  The full composition is provided below for both dataset versions, the downsampled one and the full variant:
23
- - `domain-pool`: 5,671,880 domains, labelled across three axes, all with at least one categorical label that can pertain to its reliability (e.g., 'fake news' or 'adult content').
24
  Categories are listed below.
25
  - `domain-pool-downsampled`: 149,086 domains, where the dominating categories (ones with more than 20,000 datapoints) are downsampled to 15,000 or less.
26
  Due to overlaps between datasets, this brings some of the dominating categories to counts between 10 and 15,000; the processing includes an iterative optimizer that tries to minimize such loss.
27
- This is beneficial because the entire domain pool is predominantly composed of a few large categories (e.g., adult content accounts for more than 4 million domains).
28
 
29
 
30
- Some of the primary contributors to the dataset are:
31
- - [UT1](http://dsi.ut-capitole.fr/blacklists/index_en.php) by the University of Toulouse Capitole (88.6%),
32
- - [DQR](https://academic.oup.com/pnasnexus/article/2/9/pgad286/7258994?login=false) by Lin et al. (7.6%),
33
- - Wikipedia (3.6%),
34
- - [Lasser et al.]()'s data (3.1%).
35
 
36
 
37
 
@@ -59,8 +50,8 @@ More precisely,
59
  <!-- - 25th perc.: 0.44, -->
60
  <!-- - median: 0.64, -->
61
  - mean: 0.59,
62
- <!-- - 75th perc. = 0.75,
63
- - max = 1.00 -->
64
 
65
  Distribution:
66
 
@@ -74,99 +65,79 @@ Distribution:
74
 
75
 
76
 
77
- ##### Reliability 3-class
78
 
79
  ![Value Distribution](https://cdn-uploads.huggingface.co/production/uploads/681e3663829118a837bbaeb3/nOI3VmV_xedousOtg2F9c.png)
80
 
81
- | Value | Domains |
82
- |--------|---------|
83
- | low | 6440 |
84
- | high | 5426 |
85
- | medium | 309 |
86
 
87
- #### Factuality
88
 
89
- | Factuality | Count |
90
- | ---- | ---- |
91
- | Very High | 96 |
92
- | High | 3,897
93
- | Medium | 5,853
94
- | Low | 2,086
95
- | Very Low | 230 |
96
 
97
- #### Bias
98
 
 
99
 
 
 
 
 
 
 
 
 
100
 
101
- | Bias Category | Count |
102
- |----------------|---------|
103
- | float | 11,378 |
104
- | far-right | 270 |
105
- | right | 483 |
106
- | right-center | 963 |
107
- | conspiracy | 201 |
108
- | pseudoscience | 254 |
109
- | least biased | 951 |
110
- | pro-science | 109 |
111
- | left-center | 729 |
112
- | left | 301 |
113
- | far-left | 23 |
114
 
115
 
116
- ### Domain Composition
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
117
 
118
- | dataset_domain | domain_count |
119
- |---------------------------|-------------:|
120
- | political | 17,025 |
121
- | adult | 15,000 |
122
- | phishing | 15,000 |
123
- | gambling | 14,993 |
124
- | shopping | 14,932 |
125
- | cryptojacking | 14,916 |
126
- | games | 14,887 |
127
- | jobsearch | 13,838 |
128
- | malware | 13,766 |
129
- | bank | 6,311 |
130
- | dating | 6,265 |
131
- | vpn | 6,028 |
132
- | press | 4,486 |
133
- | publicite | 4,414 |
134
- | audio-video | 3,465 |
135
- | sports | 2,292 |
136
- | coordinated campaigns | 2,250 |
137
- | blog & forums | 1,656 |
138
- | bitcoin | 1,291 |
139
- | filehosting | 819 |
140
- | manga | 655 |
141
- | social_networks | 654 |
142
- | drogue | 578 |
143
- | celebrity | 570 |
144
- | radio | 547 |
145
- | stalkerware | 517 |
146
- | educational | 457 |
147
- | financial | 452 |
148
- | webmail | 407 |
149
- | agressif | 278 |
150
- | chat | 193 |
151
- | translation | 171 |
152
- | lingerie | 165 |
153
- | legal | 154 |
154
- | health | 148 |
155
- | cult | 142 |
156
- | marketingware | 77 |
157
- | ai | 73 |
158
- | child | 70 |
159
- | cleaning | 63 |
160
- | mobile-phone | 48 |
161
- | dangerous_material | 38 |
162
- | cooking | 37 |
163
- | astrology | 28 |
164
- | sexual_education | 17 |
165
- | educational_games | 9 |
166
- | religious associations | 1 |
167
- | special | 1 |
168
 
169
  ### Data sources
 
 
 
 
 
 
 
 
 
 
170
  ```
171
  ── SAMPLED POOL: 148,830 domains ──
172
  ut1 131,795 ( 88.6%)
@@ -196,6 +167,119 @@ Distribution:
196
  edmo_hubs 16 ( 0.0%)
197
  ```
198
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
199
  ## `domain-pool` (full)
200
 
201
  The full dataset, with no downsampling, has a majority of datapoints sourced from ut1 (96.1%) and phishing datasets (1.8%, 0.8% and 0.7% for `url-phish`, `phish-dataset` and `legit-phish`
@@ -204,6 +288,8 @@ respectively). It is overwhelmingly composed of adult websites (4.6 mio), phishi
204
 
205
  ### Domain Composition
206
 
 
 
207
  | dataset_domain | domain_count |
208
  |---------------------------|-------------:|
209
  | adult | 4,592,820 |
@@ -295,6 +381,60 @@ respectively). It is overwhelmingly composed of adult websites (4.6 mio), phishi
295
  Hub No. EP/Y028872/1*. This research was also enabled in part by compute resources provided by Mila (mila.quebec) and Compute Canada.
296
 
297
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
298
 
299
  <!-- pool.csv: 5671880
300
  downsampled.csv: 149086
@@ -321,232 +461,3 @@ low 6440
321
  high 5426
322
  medium 309 -->
323
 
324
- Reliability categories
325
- ----------------------
326
- value domains
327
- adult 15000
328
- gambling 14993
329
- shopping 14939
330
- cryptojacking 14918
331
- games 14905
332
- jobsearch 13752
333
- phishing 13586
334
- malware 13553
335
- bank 6316
336
- dating 6268
337
- vpn 6030
338
- press 4524
339
- publicite 4424
340
- audio-video 3475
341
- sports 2295
342
- coord 1662
343
- blog & forums 1654
344
- non-phishing 1418
345
- fake news 1398
346
- bitcoin 1280
347
- political misinformation 1094
348
- filehosting 823
349
- manga 652
350
- social_networks 651
351
- impersonating 586
352
- drogue 583
353
- celebrity 565
354
- radio 549
355
- stalkerware 517
356
- financial 454
357
- webmail 406
358
- dicts 279
359
- agressif 278
360
- tools 198
361
- chat 193
362
- Fake news 191
363
- translation 176
364
- legal 171
365
- lingerie 162
366
- health 155
367
- cult 141
368
- Other networks 117
369
- marketingware 77
370
- Pseudoscience and junk science 72
371
- ai 72
372
- child 70
373
- cleaning 67
374
- Generative AI 65
375
- Imposter site 49
376
- mobile-phone 47
377
- hate speech 45
378
- dangerous_material 41
379
- cooking 37
380
- Parody site 36
381
- fact-checkers 35
382
- astrology 27
383
- Fraudulent fact-checking websites 21
384
- Hate groups 18
385
- sexual_education 17
386
- educational_games 8
387
- hyper-partisan 5
388
- Fraudulent virtual encyclopedias 2
389
- Global 2
390
- religious associations 1
391
- special 1
392
-
393
- Factuality categories
394
- ---------------------
395
- value domains
396
- medium 5889
397
- high 3962
398
- low 2089
399
- very low 230
400
- very high 103
401
-
402
- Bias continuous
403
- ---------------
404
- count=11477 min=0.2625 p25=0.5043 median=0.6553 mean=0.6454 p75=0.7696 max=0.9988
405
- range domains
406
- [0.2, 0.3) 4
407
- [0.3, 0.4) 504
408
- [0.4, 0.5) 2327
409
- [0.5, 0.6) 1592
410
- [0.6, 0.7) 2384
411
- [0.7, 0.8) 2549
412
- [0.8, 0.9) 1867
413
- [0.9, 1.0] 250
414
-
415
- Bias categories
416
- ---------------
417
- value domains
418
- right-center 969
419
- least biased 966
420
- left-center 757
421
- right 483
422
- left 305
423
- far-right 270
424
- pseudoscience 256
425
- conspiracy 202
426
- pro-science 118
427
- far-left 23
428
- pro- science 1
429
-
430
-
431
-
432
- Target values
433
- -------------
434
- value domains
435
- USA 3640
436
- Czech Republic 360
437
- India 348
438
- Europe 317
439
- China 241
440
- Global 221
441
- United Kingdom 190
442
- Canada 183
443
- North Macedonia 178
444
- Myanmar 97
445
- Iran 94
446
- Ghana 82
447
- Ukraine 58
448
- Australia 46
449
- Georgia 45
450
- France 38
451
- Hong Kong 36
452
- Israel 34
453
- Russia 34
454
- Germany 28
455
- Africa 27
456
- Japan 26
457
- Italy 24
458
- South Korea 24
459
- Turkey 21
460
- South Africa 20
461
- Cambodia 17
462
- Taiwan 17
463
- Central African Republic 16
464
- Netherlands 16
465
- Pakistan 14
466
- Spain 14
467
- Sweden 14
468
- United Arab Emirates 14
469
- Switzerland 13
470
- Brazil 12
471
- Ireland 12
472
- Egypt 10
473
- Mexico 10
474
- Romania 10
475
- Kosovo 9
476
- Philippines 9
477
- Tunisia 9
478
- Argentina 8
479
- Austria 8
480
- Belgium 8
481
- Nigeria 8
482
- Poland 7
483
- Bangladesh 6
484
- Ecuador 6
485
- Greece 6
486
- South Asia 6
487
- Cyprus 5
488
- Denmark 5
489
- Finland 5
490
- Indonesia 5
491
- Malaysia 5
492
- Venezuela 5
493
- Bulgaria 4
494
- Kenya 4
495
- Norway 4
496
- Oceania 4
497
- Saudi Arabia 4
498
- Thailand 4
499
- Algeria 3
500
- New Zealand 3
501
- Serbia 3
502
- Singapore 3
503
- Tanzania 3
504
- Albania 2
505
- Armenia 2
506
- Chile 2
507
- Colombia 2
508
- Croatia 2
509
- Iceland 2
510
- Iraq 2
511
- Jordan 2
512
- Lebanon 2
513
- Lithuania 2
514
- North Korea 2
515
- Portugal 2
516
- Slovenia 2
517
- Sri Lanka 2
518
- Andorra 1
519
- Belarus 1
520
- Beligium 1
521
- Bosnia and Herzegovina 1
522
- Cameroon 1
523
- Costa Rica 1
524
- Cuba 1
525
- Estonia 1
526
- Guam 1
527
- Guinea 1
528
- Hungary 1
529
- Latvia 1
530
- Luxembourg 1
531
- Morocco 1
532
- Puerto Rico 1
533
- Qatar 1
534
- Syria 1
535
- Uruguay 1
536
- Zimbabwe 1
537
-
538
- Perpetrator values
539
- ------------------
540
- value domains
541
- Russia 551
542
- China 206
543
- Israel 143
544
- India 52
545
- Hong Kong 34
546
- Turkey 21
547
- Iran 18
548
- Europe 3
549
- Lebanon 2
550
- Thailand 2
551
- Benin 1
552
- USA 1
 
2
  license: cc-by-4.0
3
  ---
4
 
 
 
 
 
5
  # Dataset Card for *`domain-pool`*
6
 
7
 
8
  `domain-pool` is a fine grained and cross-domain aggregate labelled set of web domains. Its default form has majority categories downsampled to present a more balanced set of
9
  148,830 domains, while the full imbalanced set has 5,671,355 datapoints.
10
  These web domains are mainly labelled across three axes:
11
+ - reliability labels form the principal category (spans all datapoints). Reliability scores can be numerical, then normalized to [0.0,1.0] where higher is better; or categorical. Categories are listed below.
12
+ - factuality labels are sparser, spanning around 12k domains. They are categorical: a domain can have low, medium or high factuality, as assessed by sources like fact-checking organisations.
13
  - bias labels can, like the reliability ones, either be a continuous score on [0.0,1.0], or a category (can either be quantitative or across left/right axis).
14
  All domains also have the original data source indicated, and the dataset's domain scope (e.g., misinformation, or malware; domain-pool spans Y domains).
15
  A large part of these data sources are open-sources academic datasets, and the labels for Y domains were also collected manually
16
  from online sources (governmental, journalistic or academic) that gathered domain lists in non-machine readable format.
17
 
18
  The full composition is provided below for both dataset versions, the downsampled one and the full variant:
19
+ - `domain-pool`: 5,671,880 domains, labelled across the three axes, all with at least one categorical label that can pertain to its reliability (e.g., 'fake news' or 'adult content').
20
  Categories are listed below.
21
  - `domain-pool-downsampled`: 149,086 domains, where the dominating categories (ones with more than 20,000 datapoints) are downsampled to 15,000 or less.
22
  Due to overlaps between datasets, this brings some of the dominating categories to counts between 10 and 15,000; the processing includes an iterative optimizer that tries to minimize such loss.
23
+ Downsampling here is beneficial because the entire domain pool is predominantly composed of a few large categories (e.g., adult content accounts for more than 4 million domains).
24
 
25
 
 
 
 
 
 
26
 
27
 
28
 
 
50
  <!-- - 25th perc.: 0.44, -->
51
  <!-- - median: 0.64, -->
52
  - mean: 0.59,
53
+ <!-- - 75th perc. = 0.75, - max = 1.00 -->
54
+
55
 
56
  Distribution:
57
 
 
65
 
66
 
67
 
68
+ ##### Reliability (3-class)
69
 
70
  ![Value Distribution](https://cdn-uploads.huggingface.co/production/uploads/681e3663829118a837bbaeb3/nOI3VmV_xedousOtg2F9c.png)
71
 
72
+ | Value | low | high | medium |
73
+ |---------|------|------|--------|
74
+ | Domains | 6440 | 5426 | 309 |
 
 
75
 
 
76
 
77
+ ##### Reliability (categorical)
 
 
 
 
 
 
78
 
79
+ ![Domain Distribution](https://cdn-uploads.huggingface.co/production/uploads/681e3663829118a837bbaeb3/ha-Wn2jgyfBBNG1BeojSH.png)
80
 
81
+ These are the 5 largest categories; the full list can be found at the bottom of the dataset card
82
 
83
+ | Domain / area | Count |
84
+ |---------------------------|--------------|
85
+ | political | 17,180 |
86
+ | adult | 15,000 |
87
+ | phishing | 15,000 |
88
+ | gambling | 14,993 |
89
+ | shopping | 14,939 |
90
+ | cryptojacking, games, jobsearch, malware... | 90k+ |
91
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92
 
93
 
94
+ ##### Factuality
95
+
96
+
97
+ ![Value Distribution](https://cdn-uploads.huggingface.co/production/uploads/681e3663829118a837bbaeb3/NI7Qgl_QkXaGTdqz2-vot.png)
98
+
99
+ | Factuality | very low | low | medium | high | very high |
100
+ |-------------|--------|------|------|----------|-----------|
101
+ | Count | 230 | 2089 | 5889 | 3962 | 103 |
102
+
103
+
104
+ ##### Bias (continuous)
105
+
106
+ - Count: 11,477
107
+ - Mean: 0.65
108
+ <!-- count=11477 min=0.2625 p25=0.5043 median=0.6553 mean=0.6454 p75=0.7696 max=0.9988 -->
109
+
110
+
111
+
112
+ ![Bias Distribution](https://cdn-uploads.huggingface.co/production/uploads/681e3663829118a837bbaeb3/FHR_Ri92DuFhFauVbuLbx.png)
113
+
114
+ | Range | [0.2, 0.3) | [0.3, 0.4) | [0.4, 0.5) | [0.5, 0.6) | [0.6, 0.7) | [0.7, 0.8) | [0.8, 0.9) | [0.9, 1.0] |
115
+ |--------------|------------|------------|------------|------------|------------|------------|------------|------------|
116
+ | Domains | 4 | 504 | 2327 | 1592 | 2384 | 2549 | 1867 | 250 |
117
+
118
+
119
+
120
+
121
+ ##### Bias (categorical)
122
+
123
+ | Bias Category | Far-Left | Left | Left-Center | Least Biased | Right-Center | Right | Far-Right | Pro-Science | Pseudoscience | Conspiracy |
124
+ |---------------|----------|------|-------------|--------------|--------------|-------|-----------|-------------|---------------|------------|
125
+ | Domains | 23 | 305 | 757 | 966 | 969 | 483 | 270 | 118 | 256 | 202 |
126
+
127
+
128
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
129
 
130
  ### Data sources
131
+
132
+ Some of the primary contributors to the dataset are:
133
+ - [UT1](http://dsi.ut-capitole.fr/blacklists/index_en.php) by the University of Toulouse Capitole (88.6%),
134
+ - [DQR](https://academic.oup.com/pnasnexus/article/2/9/pgad286/7258994?login=false) by Lin et al. (7.6%),
135
+ - Wikipedia (3.6%),
136
+ - [Lasser et al.]()'s data (3.1%).
137
+
138
+
139
+ The full list:
140
+
141
  ```
142
  ── SAMPLED POOL: 148,830 domains ──
143
  ut1 131,795 ( 88.6%)
 
167
  edmo_hubs 16 ( 0.0%)
168
  ```
169
 
170
+ ### Geographical Distribution
171
+
172
+ Political-scoped data sources largely have country attribution. In most cases, it's the country or region that the misinformation / propaganda
173
+ targets. In the case of coordinated disinformation campaigns, the perpetrators may also be attributed:
174
+
175
+
176
+
177
+ | Country | Target | Perp |
178
+ |----------------------------|--------|------|
179
+ | USA | 3640 | 1 |
180
+ | Czech Republic | 360 | 0 |
181
+ | India | 348 | 52 |
182
+ | Europe | 317 | 3 |
183
+ | China | 241 | 206 |
184
+ | Global | 221 | 0 |
185
+ | United Kingdom | 190 | 0 |
186
+ | Canada | 183 | 0 |
187
+ | North Macedonia | 178 | 0 |
188
+ | Myanmar | 97 | 0 |
189
+ | Iran | 94 | 18 |
190
+ | Ghana | 82 | 0 |
191
+ | Ukraine | 58 | 0 |
192
+ | Australia | 46 | 0 |
193
+ | Georgia | 45 | 0 |
194
+ | France | 38 | 0 |
195
+ | Hong Kong | 36 | 34 |
196
+ | Israel | 34 | 143 |
197
+ | Russia | 34 | 551 |
198
+ | Germany | 28 | 0 |
199
+ | Africa | 27 | 0 |
200
+ | Japan | 26 | 0 |
201
+ | Italy | 24 | 0 |
202
+ | South Korea | 24 | 0 |
203
+ | Turkey | 21 | 21 |
204
+ | South Africa | 20 | 0 |
205
+ | Cambodia | 17 | 0 |
206
+ | Taiwan | 17 | 0 |
207
+ | Central African Republic | 16 | 0 |
208
+ | Netherlands | 16 | 0 |
209
+ | Pakistan | 14 | 0 |
210
+ | Spain | 14 | 0 |
211
+ | Sweden | 14 | 0 |
212
+ | United Arab Emirates | 14 | 0 |
213
+ | Switzerland | 13 | 0 |
214
+ | Brazil | 12 | 0 |
215
+ | Ireland | 12 | 0 |
216
+ | Egypt | 10 | 0 |
217
+ | Mexico | 10 | 0 |
218
+ | Romania | 10 | 0 |
219
+ | Kosovo | 9 | 0 |
220
+ | Philippines | 9 | 0 |
221
+ | Tunisia | 9 | 0 |
222
+ | Argentina | 8 | 0 |
223
+ | Austria | 8 | 0 |
224
+ | Belgium | 8 | 0 |
225
+ | Nigeria | 8 | 0 |
226
+ | Poland | 7 | 0 |
227
+ | Bangladesh | 6 | 0 |
228
+ | Ecuador | 6 | 0 |
229
+ | Greece | 6 | 0 |
230
+ | South Asia | 6 | 0 |
231
+ | Cyprus | 5 | 0 |
232
+ | Denmark | 5 | 0 |
233
+ | Finland | 5 | 0 |
234
+ | Indonesia | 5 | 0 |
235
+ | Malaysia | 5 | 0 |
236
+ | Venezuela | 5 | 0 |
237
+ | Bulgaria | 4 | 0 |
238
+ | Kenya | 4 | 0 |
239
+ | Norway | 4 | 0 |
240
+ | Oceania | 4 | 0 |
241
+ | Saudi Arabia | 4 | 0 |
242
+ | Thailand | 4 | 2 |
243
+ | Algeria | 3 | 0 |
244
+ | New Zealand | 3 | 0 |
245
+ | Serbia | 3 | 0 |
246
+ | Singapore | 3 | 0 |
247
+ | Tanzania | 3 | 0 |
248
+ | Albania | 2 | 0 |
249
+ | Armenia | 2 | 0 |
250
+ | Chile | 2 | 0 |
251
+ | Colombia | 2 | 0 |
252
+ | Croatia | 2 | 0 |
253
+ | Iceland | 2 | 0 |
254
+ | Iraq | 2 | 0 |
255
+ | Jordan | 2 | 0 |
256
+ | Lebanon | 2 | 2 |
257
+ | Lithuania | 2 | 0 |
258
+ | North Korea | 2 | 0 |
259
+ | Portugal | 2 | 0 |
260
+ | Slovenia | 2 | 0 |
261
+ | Sri Lanka | 2 | 0 |
262
+ | Andorra | 1 | 0 |
263
+ | Belarus | 1 | 0 |
264
+ | Beligium | 1 | 0 |
265
+ | Bosnia and Herzegovina | 1 | 0 |
266
+ | Cameroon | 1 | 0 |
267
+ | Costa Rica | 1 | 0 |
268
+ | Cuba | 1 | 0 |
269
+ | Estonia | 1 | 0 |
270
+ | Guam | 1 | 0 |
271
+ | Guinea | 1 | 0 |
272
+ | Hungary | 1 | 0 |
273
+ | Latvia | 1 | 0 |
274
+ | Luxembourg | 1 | 0 |
275
+ | Morocco | 1 | 0 |
276
+ | Puerto Rico | 1 | 0 |
277
+ | Qatar | 1 | 0 |
278
+ | Syria | 1 | 0 |
279
+ | Uruguay | 1 | 0 |
280
+ | Zimbabwe | 1 | 0 |
281
+ | Benin | 0 | 1 |
282
+
283
  ## `domain-pool` (full)
284
 
285
  The full dataset, with no downsampling, has a majority of datapoints sourced from ut1 (96.1%) and phishing datasets (1.8%, 0.8% and 0.7% for `url-phish`, `phish-dataset` and `legit-phish`
 
288
 
289
  ### Domain Composition
290
 
291
+ ![Domain Distribution2](https://cdn-uploads.huggingface.co/production/uploads/681e3663829118a837bbaeb3/4Wc91wjxEGtTpWY0K0BMr.png)
292
+
293
  | dataset_domain | domain_count |
294
  |---------------------------|-------------:|
295
  | adult | 4,592,820 |
 
381
  Hub No. EP/Y028872/1*. This research was also enabled in part by compute resources provided by Mila (mila.quebec) and Compute Canada.
382
 
383
 
384
+ ### Full category counts:
385
+
386
+ #### Downsampled:
387
+
388
+ | dataset_domain | domain_count |
389
+ |----------------------------|--------------|
390
+ | political misinformation | 17180 |
391
+ | adult | 15000 |
392
+ | phishing | 15000 |
393
+ | gambling | 14993 |
394
+ | shopping | 14939 |
395
+ | cryptojacking | 14918 |
396
+ | games | 14905 |
397
+ | jobsearch | 13752 |
398
+ | malware | 13680 |
399
+ | bank | 6316 |
400
+ | dating | 6268 |
401
+ | vpn | 6030 |
402
+ | press | 4524 |
403
+ | publicite | 4424 |
404
+ | audio-video | 3475 |
405
+ | sports | 2295 |
406
+ | coordinated campaigns | 2248 |
407
+ | blog & forums | 1654 |
408
+ | bitcoin | 1280 |
409
+ | filehosting | 823 |
410
+ | manga | 652 |
411
+ | social_networks | 651 |
412
+ | drugs | 583 |
413
+ | celebrity | 565 |
414
+ | radio | 549 |
415
+ | stalkerware | 517 |
416
+ | educational | 477 |
417
+ | financial | 454 |
418
+ | webmail | 406 |
419
+ | agressif | 278 |
420
+ | chat | 193 |
421
+ | translation | 176 |
422
+ | legal | 171 |
423
+ | lingerie | 162 |
424
+ | health | 155 |
425
+ | cult | 141 |
426
+ | marketingware | 77 |
427
+ | ai | 72 |
428
+ | child | 70 |
429
+ | cleaning | 67 |
430
+ | mobile-phone | 47 |
431
+ | dangerous_material | 41 |
432
+ | cooking | 37 |
433
+ | astrology | 27 |
434
+ | sexual_education | 17 |
435
+ | educational_games | 8 |
436
+ | religious associations | 1 |
437
+ | special | 1 |
438
 
439
  <!-- pool.csv: 5671880
440
  downsampled.csv: 149086
 
461
  high 5426
462
  medium 309 -->
463