Update README.md
Browse files
README.md
CHANGED
@@ -34,4 +34,36 @@ language:
|
|
34 |
pretty_name: Comments under Le Monde Ukraine war articles (1 year)
|
35 |
size_categories:
|
36 |
- 100K<n<1M
|
37 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
pretty_name: Comments under Le Monde Ukraine war articles (1 year)
|
35 |
size_categories:
|
36 |
- 100K<n<1M
|
37 |
+
---
|
38 |
+
|
39 |
+
## Comments under Le Monde Ukraine War Articles (1 Year)
|
40 |
+
|
41 |
+
### Description
|
42 |
+
This dataset contains 175k comments extracted from Le Monde articles about the Ukraine war during its first year (February 2022 to 2023).
|
43 |
+
Among these, around 500 comments are manually labeled into categories: 0. Explicit support for Ukraine, 1. pro Russia, 2. "Other".
|
44 |
+
|
45 |
+
### Dataset Structure
|
46 |
+
|
47 |
+
#### Features
|
48 |
+
- `text`: The comment text (string).
|
49 |
+
- `label`: The label for the comment (integer). The labels are as follows:
|
50 |
+
- 0: pro_Ukraine
|
51 |
+
- 1: pro_Russia
|
52 |
+
- 2: other
|
53 |
+
- 4: no_label (the unlabeled data).
|
54 |
+
|
55 |
+
#### Splits
|
56 |
+
Train and validation are manually labeled. Unlabeled data could be used for knowledge distillation for instance.
|
57 |
+
- `train`: 323 examples.
|
58 |
+
- `validation`: 139 examples.
|
59 |
+
- `unlabeled`: 174,891 examples.
|
60 |
+
|
61 |
+
|
62 |
+
### Additional Information
|
63 |
+
|
64 |
+
- **Homepage**: [Project Repository](https://github.com/matthieuvion/lmd_classi)
|
65 |
+
- **License**: MIT License
|
66 |
+
- **Language**: French
|
67 |
+
- **Task Categories**: Text Classification
|
68 |
+
- **Size Categories**: 100K < n < 1M
|
69 |
+
|