Datasets:
dainis-boumber
commited on
Commit
•
d7450fa
1
Parent(s):
3b77f86
redone yaml following GLUE
Browse files
README.md
CHANGED
@@ -1,74 +1,154 @@
|
|
1 |
---
|
2 |
-
|
3 |
-
-
|
4 |
-
|
5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
path: fake_news/train.jsonl
|
7 |
-
-
|
8 |
path: fake_news/test.jsonl
|
9 |
-
-
|
10 |
path: fake_news/validation.jsonl
|
11 |
-
-
|
12 |
-
|
13 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
path: job_scams/train.jsonl
|
15 |
-
-
|
16 |
path: job_scams/test.jsonl
|
17 |
-
-
|
18 |
path: job_scams/validation.jsonl
|
19 |
-
-
|
20 |
-
|
21 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
path: phishing/train.jsonl
|
23 |
-
-
|
24 |
path: phishing/test.jsonl
|
25 |
-
-
|
26 |
path: phishing/validation.jsonl
|
27 |
-
-
|
28 |
-
|
29 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
path: political_statements/train.jsonl
|
31 |
-
-
|
32 |
path: political_statements/test.jsonl
|
33 |
-
-
|
34 |
path: political_statements/validation.jsonl
|
35 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
data_files:
|
37 |
-
-
|
38 |
path: product_reviews/train.jsonl
|
39 |
-
-
|
40 |
path: product_reviews/test.jsonl
|
41 |
-
-
|
42 |
path: product_reviews/validation.jsonl
|
43 |
-
-
|
44 |
-
|
45 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
path: sms/train.jsonl
|
47 |
-
-
|
48 |
path: sms/test.jsonl
|
49 |
-
-
|
50 |
path: sms/validation.jsonl
|
51 |
-
-
|
52 |
-
|
53 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
path: twitter_rumours/train.jsonl
|
55 |
-
-
|
56 |
path: twitter_rumours/test.jsonl
|
57 |
-
-
|
58 |
path: twitter_rumours/validation.jsonl
|
59 |
-
|
60 |
-
|
61 |
-
-
|
62 |
-
|
63 |
-
-
|
64 |
-
|
65 |
-
-
|
|
|
66 |
---
|
67 |
|
68 |
# GDDs-2.0
|
69 |
|
70 |
-
The Generalized Deception Dataset version 2.0 is a labeled corpus containing over 95000 samples of
|
71 |
-
deceptive and truthful texts from a number of independent domains and tasks.
|
72 |
|
73 |
## Authors
|
74 |
|
@@ -342,4 +422,4 @@ https://figshare.com/articles/dataset/PHEME_dataset_of_rumours_and_non-rumours/4
|
|
342 |
was used in creation of this dataset. We took source tweets only, and ignored replies to them.
|
343 |
We used source tweet's label as being a rumour or non-rumour to label it as deceptive or non-deceptive.
|
344 |
|
345 |
-
The training set contains 4631 samples, the validation and the test sets have 579 samples each
|
|
|
1 |
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
license:
|
5 |
+
- mit
|
6 |
+
size_categories:
|
7 |
+
- 10K<n<100K
|
8 |
+
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 |
+
dataset_info:
|
20 |
+
- config: fake_news
|
21 |
+
features:
|
22 |
+
- name: text
|
23 |
+
dtype: string
|
24 |
+
- name: label
|
25 |
+
dtype:
|
26 |
+
class_label:
|
27 |
+
names:
|
28 |
+
'0': non-deceptive
|
29 |
+
'1': deceptive
|
30 |
+
splits:
|
31 |
+
- name: train
|
32 |
path: fake_news/train.jsonl
|
33 |
+
- name: test
|
34 |
path: fake_news/test.jsonl
|
35 |
+
- name: validation
|
36 |
path: fake_news/validation.jsonl
|
37 |
+
- config: job_scams
|
38 |
+
features:
|
39 |
+
- name: text
|
40 |
+
dtype: string
|
41 |
+
- name: label
|
42 |
+
dtype:
|
43 |
+
class_label:
|
44 |
+
names:
|
45 |
+
'0': non-deceptive
|
46 |
+
'1': deceptive
|
47 |
+
splits:
|
48 |
+
- name: train
|
49 |
path: job_scams/train.jsonl
|
50 |
+
- name: test
|
51 |
path: job_scams/test.jsonl
|
52 |
+
- name: validation
|
53 |
path: job_scams/validation.jsonl
|
54 |
+
- config: phishing
|
55 |
+
features:
|
56 |
+
- name: text
|
57 |
+
dtype: string
|
58 |
+
- name: label
|
59 |
+
dtype:
|
60 |
+
class_label:
|
61 |
+
names:
|
62 |
+
'0': non-deceptive
|
63 |
+
'1': deceptive
|
64 |
+
splits:
|
65 |
+
- name: train
|
66 |
path: phishing/train.jsonl
|
67 |
+
- name: test
|
68 |
path: phishing/test.jsonl
|
69 |
+
- name: validation
|
70 |
path: phishing/validation.jsonl
|
71 |
+
- config: political_statements
|
72 |
+
features:
|
73 |
+
- name: text
|
74 |
+
dtype: string
|
75 |
+
- name: label
|
76 |
+
dtype:
|
77 |
+
class_label:
|
78 |
+
names:
|
79 |
+
'0': non-deceptive
|
80 |
+
'1': deceptive
|
81 |
+
splits:
|
82 |
+
- name: train
|
83 |
path: political_statements/train.jsonl
|
84 |
+
- name: test
|
85 |
path: political_statements/test.jsonl
|
86 |
+
- name: validation
|
87 |
path: political_statements/validation.jsonl
|
88 |
+
- config: product_reviews
|
89 |
+
features:
|
90 |
+
- name: text
|
91 |
+
dtype: string
|
92 |
+
- name: label
|
93 |
+
dtype:
|
94 |
+
class_label:
|
95 |
+
names:
|
96 |
+
'0': non-deceptive
|
97 |
+
'1': deceptive
|
98 |
data_files:
|
99 |
+
- name: train
|
100 |
path: product_reviews/train.jsonl
|
101 |
+
- name: test
|
102 |
path: product_reviews/test.jsonl
|
103 |
+
- name: validation
|
104 |
path: product_reviews/validation.jsonl
|
105 |
+
- config: sms
|
106 |
+
features:
|
107 |
+
- name: text
|
108 |
+
dtype: string
|
109 |
+
- name: label
|
110 |
+
dtype:
|
111 |
+
class_label:
|
112 |
+
names:
|
113 |
+
'0': non-deceptive
|
114 |
+
'1': deceptive
|
115 |
+
splits:
|
116 |
+
- name: train
|
117 |
path: sms/train.jsonl
|
118 |
+
- name: test
|
119 |
path: sms/test.jsonl
|
120 |
+
- name: validation
|
121 |
path: sms/validation.jsonl
|
122 |
+
- config: twitter_rumours
|
123 |
+
features:
|
124 |
+
- name: text
|
125 |
+
dtype: string
|
126 |
+
- name: label
|
127 |
+
dtype:
|
128 |
+
class_label:
|
129 |
+
names:
|
130 |
+
'0': non-deceptive
|
131 |
+
'1': deceptive
|
132 |
+
splits:
|
133 |
+
- name: train
|
134 |
path: twitter_rumours/train.jsonl
|
135 |
+
- name: test
|
136 |
path: twitter_rumours/test.jsonl
|
137 |
+
- name: validation
|
138 |
path: twitter_rumours/validation.jsonl
|
139 |
+
config_names:
|
140 |
+
- fake_news
|
141 |
+
- job_scams
|
142 |
+
- phishing
|
143 |
+
- political_statements
|
144 |
+
- product_reviews
|
145 |
+
- sms
|
146 |
+
- twitter_rumours
|
147 |
---
|
148 |
|
149 |
# GDDs-2.0
|
150 |
|
151 |
+
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.
|
|
|
152 |
|
153 |
## Authors
|
154 |
|
|
|
422 |
was used in creation of this dataset. We took source tweets only, and ignored replies to them.
|
423 |
We used source tweet's label as being a rumour or non-rumour to label it as deceptive or non-deceptive.
|
424 |
|
425 |
+
The training set contains 4631 samples, the validation and the test sets have 579 samples each
|