Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -1,28 +1,9 @@
|
|
1 |
-
Dataset for training an entailment classifier to recognize approval/disapproval of politicians.
|
2 |
-
Documents are Tweets from Kawintiranon (2022), the MTSD dataset, as well as Tweets and sentences taken weekly newsletters for select politicians from the 115th, 116th, and 117th congress.
|
3 |
-
Documents are triple coded -- once from the original compilers of the dataset, once from GPT-4, and a third time to adjudicate discrepancies between the two.
|
4 |
-
|
5 |
-
Twitter handles from politicians in the dataset have been replaced by their name. Be aware that "rt @realdonaldtrump You're a liar!" has been replaced with "rt trump You're a liar!",
|
6 |
-
And means the author is retweeting Trump calling someone else a liar, not the author calling Trump a liar.
|
7 |
-
|
8 |
-
Stance:
|
9 |
-
-1: Against: The document is critical of the target.
|
10 |
-
0: Neutral: The document doesn't express an opinion about the target or it can't be determined what the expressed opinion is with the given context.
|
11 |
-
1: Support: The document expresses support for the target. Expressing collaboration on bills or letters is considered support.
|
12 |
-
|
13 |
-
Label:
|
14 |
-
0: Entail
|
15 |
-
1: Not Entail
|
16 |
-
|
17 |
-
In addition to 1,000 documents about targets in the training set, the test set contains documents about 6 politicians not included in the training or validation data:
|
18 |
-
- Ted Cruz
|
19 |
-
- Hakeem Jeffries
|
20 |
-
- Madison Cawthorn
|
21 |
-
- Alexandria Ocasio-Cortez
|
22 |
-
- Mitt Romney
|
23 |
-
- Kyrsten Sinema
|
24 |
-
|
25 |
---
|
|
|
|
|
|
|
|
|
|
|
26 |
license: mit
|
27 |
configs:
|
28 |
- config_name: default
|
@@ -58,3 +39,27 @@ dataset_info:
|
|
58 |
download_size: 4211025
|
59 |
dataset_size: 8353461
|
60 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
pretty name: PoliStance Affect
|
3 |
+
task_categories:
|
4 |
+
- Zero Shot Classification
|
5 |
+
- NLI
|
6 |
+
- Stance Detection
|
7 |
license: mit
|
8 |
configs:
|
9 |
- config_name: default
|
|
|
39 |
download_size: 4211025
|
40 |
dataset_size: 8353461
|
41 |
---
|
42 |
+
|
43 |
+
Dataset for training an entailment classifier to recognize approval/disapproval of politicians.
|
44 |
+
Documents are Tweets from Kawintiranon (2022), the MTSD dataset, as well as Tweets and sentences taken weekly newsletters for select politicians from the 115th, 116th, and 117th congress.
|
45 |
+
Documents are triple coded -- once from the original compilers of the dataset, once from GPT-4, and a third time to adjudicate discrepancies between the two.
|
46 |
+
|
47 |
+
Twitter handles from politicians in the dataset have been replaced by their name. Be aware that "rt @realdonaldtrump You're a liar!" has been replaced with "rt trump You're a liar!",
|
48 |
+
And means the author is retweeting Trump calling someone else a liar, not the author calling Trump a liar.
|
49 |
+
|
50 |
+
Stance:
|
51 |
+
-1: Against: The document is critical of the target.
|
52 |
+
0: Neutral: The document doesn't express an opinion about the target or it can't be determined what the expressed opinion is with the given context.
|
53 |
+
1: Support: The document expresses support for the target. Expressing collaboration on bills or letters is considered support.
|
54 |
+
|
55 |
+
Label:
|
56 |
+
0: Entail
|
57 |
+
1: Not Entail
|
58 |
+
|
59 |
+
In addition to 1,000 documents about targets in the training set, the test set contains documents about 6 politicians not included in the training or validation data:
|
60 |
+
- Ted Cruz
|
61 |
+
- Hakeem Jeffries
|
62 |
+
- Madison Cawthorn
|
63 |
+
- Alexandria Ocasio-Cortez
|
64 |
+
- Mitt Romney
|
65 |
+
- Kyrsten Sinema
|