Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
parquet
Sub-tasks:
multi-label-classification
Languages:
English
Size:
10K - 100K
License:
Update README.md
Browse files
README.md
CHANGED
@@ -22,3 +22,17 @@ configs:
|
|
22 |
- split: test
|
23 |
path: data/test-*
|
24 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
- split: test
|
23 |
path: data/test-*
|
24 |
---
|
25 |
+
This dataset was made specifically for multilabel classification using the following process:
|
26 |
+
|
27 |
+
1. Downloading https://mtgjson.com/api/v5/AtomicCards.json.bz2 on January 10, 2024
|
28 |
+
2. Encoding color identity of each card into the `labels` feature
|
29 |
+
```python
|
30 |
+
colors = ['B', 'G', 'R', 'U', 'W']
|
31 |
+
b = [1, 0, 0, 0, 0]
|
32 |
+
bw = [1, 0, 0, 0, 1]
|
33 |
+
gru = [0, 1, 1, 1, 0]
|
34 |
+
# and so on
|
35 |
+
```
|
36 |
+
3. Concatenating card name and card text into the `text` feature
|
37 |
+
4. `split = ds['train'].train_test_split(test_size=0.2)`
|
38 |
+
5. `split.push_to_hub("mtg-coloridentity-multilabel-classification")`
|