dainis-boumber commited on
Commit
a50cddb
1 Parent(s): 5fb457d
Files changed (1) hide show
  1. README.md +70 -65
README.md CHANGED
@@ -9,75 +9,83 @@ multilinguality:
9
  - monolingual
10
  task_categories:
11
  - text-classification
12
- pretty_name: GDDs (Generalized Deception Dataset)
 
13
  tags:
 
14
  - deception-detection
15
  - phishing
16
  - fake-news
 
17
  - opinion-spam
18
- - domain-adaptation
19
  configs:
20
- - config_name: fake_news
21
- data_files:
22
- - split: train
23
- path: fake_news/train.jsonl
24
- - split: test
25
- path: fake_news/test.jsonl
26
- - split: validation
27
- path: fake_news/validation.jsonl
28
- - config_name: job_scams
29
- data_files:
30
- - split: train
31
- path: job_scams/train.jsonl
32
- - split: test
33
- path: job_scams/test.jsonl
34
- - split: validation
35
- path: job_scams/validation.jsonl
36
- - config_name: phishing
37
- data_files:
38
- - split: train
39
- path: phishing/train.jsonl
40
- - split: test
41
- path: phishing/test.jsonl
42
- - split: validation
43
- path: phishing/validation.jsonl
44
- - config_name: political_statements
45
- data_files:
46
- - split: train
47
- path: political_statements/train.jsonl
48
- - split: test
49
- path: political_statements/test.jsonl
50
- - split: validation
51
- path: political_statements/validation.jsonl
52
- - config_name: product_reviews
53
- data_files:
54
- - split: train
55
- path: product_reviews/train.jsonl
56
- - split: test
57
- path: product_reviews/test.jsonl
58
- - split: validation
59
- path: product_reviews/validation.jsonl
60
- - config_name: sms
61
- data_files:
62
- - split: train
63
- path: sms/train.jsonl
64
- - split: test
65
- path: sms/test.jsonl
66
- - split: validation
67
- path: sms/validation.jsonl
68
- - config_name: twitter_rumours
69
- data_files:
70
- - split: train
71
- path: twitter_rumours/train.jsonl
72
- - split: test
73
- path: twitter_rumours/test.jsonl
74
- - split: validation
75
- path: twitter_rumours/validation.jsonl
76
  ---
77
 
78
- # GDDs-2.0
79
 
80
- The Generalized Deception Dataset version 2.0 is a labeled corpus containing over 95000 samples of deceptive and truthful texts from a number of independent domains and tasks.
 
 
 
 
 
81
 
82
  ## Authors
83
 
@@ -163,10 +171,10 @@ Whitespace, quotes, bulletpoints, unicode is normalized.
163
 
164
  ### Layout
165
 
166
- The directory layout of gdds is like so:
167
 
168
  ``
169
- gdds
170
  fake_news/
171
  train.jsonl
172
  test.jsonl
@@ -319,9 +327,6 @@ The training set contains 9997 samples, the validation and the test sets have 12
319
 
320
  ### PRODUCT REVIEWS
321
 
322
- We post-process and split Product Reviews dataset to ensure uniformity with Political Statements 2.0 and Twitter Rumours
323
- as they all go into form GDDS-2.0
324
-
325
  The dataset is produced from English Amazon Reviews labeled as either real or fake, relabeled as deceptive and non-deceptive respectively.
326
  The reviews cover a variety of products with no particular product dominating the dataset. Although the dataset authors filtered out
327
  non-English reviews, through outlier detection we found that the dataset still contains reviews in Spanish and other languages.
 
9
  - monolingual
10
  task_categories:
11
  - text-classification
12
+ - zero-shot-classification
13
+ pretty_name: DIFrauD - Domain-Independent Fraud Detection benchmark
14
  tags:
15
+ - fraud-detection
16
  - deception-detection
17
  - phishing
18
  - fake-news
19
+ - benchmark
20
  - opinion-spam
21
+ - multi-domain
22
  configs:
23
+ - config_name: fake_news
24
+ data_files:
25
+ - split: train
26
+ path: fake_news/train.jsonl
27
+ - split: test
28
+ path: fake_news/test.jsonl
29
+ - split: validation
30
+ path: fake_news/validation.jsonl
31
+ - config_name: job_scams
32
+ data_files:
33
+ - split: train
34
+ path: job_scams/train.jsonl
35
+ - split: test
36
+ path: job_scams/test.jsonl
37
+ - split: validation
38
+ path: job_scams/validation.jsonl
39
+ - config_name: phishing
40
+ data_files:
41
+ - split: train
42
+ path: phishing/train.jsonl
43
+ - split: test
44
+ path: phishing/test.jsonl
45
+ - split: validation
46
+ path: phishing/validation.jsonl
47
+ - config_name: political_statements
48
+ data_files:
49
+ - split: train
50
+ path: political_statements/train.jsonl
51
+ - split: test
52
+ path: political_statements/test.jsonl
53
+ - split: validation
54
+ path: political_statements/validation.jsonl
55
+ - config_name: product_reviews
56
+ data_files:
57
+ - split: train
58
+ path: product_reviews/train.jsonl
59
+ - split: test
60
+ path: product_reviews/test.jsonl
61
+ - split: validation
62
+ path: product_reviews/validation.jsonl
63
+ - config_name: sms
64
+ data_files:
65
+ - split: train
66
+ path: sms/train.jsonl
67
+ - split: test
68
+ path: sms/test.jsonl
69
+ - split: validation
70
+ path: sms/validation.jsonl
71
+ - config_name: twitter_rumours
72
+ data_files:
73
+ - split: train
74
+ path: twitter_rumours/train.jsonl
75
+ - split: test
76
+ path: twitter_rumours/test.jsonl
77
+ - split: validation
78
+ path: twitter_rumours/validation.jsonl
79
  ---
80
 
81
+ # DIFrauD - Domain Independent Fraud Detection Benchmark
82
 
83
+ Domain Independent Fraud Detection Benchmark is a labeled corpus containing over 95,854 samples of deceitful
84
+ and truthful texts from a number of independent domains and tasks. Deception, however, can be different --
85
+ in this corpus we made sure to gather strictly real examples of deception that are intentionally malicious
86
+ and cause real harm, despite them often having very little in common. Covering seven domains, this benchmark
87
+ is designed to serve as a representative slice of the various security challenges that remain open problems
88
+ today.
89
 
90
  ## Authors
91
 
 
171
 
172
  ### Layout
173
 
174
+ The directory layout of `difraud` is like so:
175
 
176
  ``
177
+ difraud
178
  fake_news/
179
  train.jsonl
180
  test.jsonl
 
327
 
328
  ### PRODUCT REVIEWS
329
 
 
 
 
330
  The dataset is produced from English Amazon Reviews labeled as either real or fake, relabeled as deceptive and non-deceptive respectively.
331
  The reviews cover a variety of products with no particular product dominating the dataset. Although the dataset authors filtered out
332
  non-English reviews, through outlier detection we found that the dataset still contains reviews in Spanish and other languages.