Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
csv
Languages:
Chinese
Size:
10K - 100K
ArXiv:
License:
Update README.md
Browse filesDescribe the dataset information, where it has abbreviated from, and how it has been modified.
README.md
CHANGED
|
@@ -1,3 +1,21 @@
|
|
| 1 |
---
|
| 2 |
license: gpl-3.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: gpl-3.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- text-classification
|
| 5 |
+
language:
|
| 6 |
+
- zh
|
| 7 |
+
size_categories:
|
| 8 |
+
- 1K<n<10K
|
| 9 |
---
|
| 10 |
+
|
| 11 |
+
This dataset is copied from [CLUE](https://github.com/CLUEbenchmark/OCNLI/tree/main/data/ocnli) with certain modifications.
|
| 12 |
+
The paper of CLUE is [OCNLI](https://arxiv.org/abs/2010.05444).
|
| 13 |
+
|
| 14 |
+
The modifications are:
|
| 15 |
+
1. Transform json file to csv file.
|
| 16 |
+
2. Encoding in UTF-8.
|
| 17 |
+
3. Remove data entries whose label value is '-'.
|
| 18 |
+
4. Replace label values, 'neutral' to 1, 'entailment' to 0, and 'contradiction' to 2.
|
| 19 |
+
|
| 20 |
+
ocnli_train_std.csv comes from train.50k.json.
|
| 21 |
+
ocnli_test_std.csv comes from dev.json.
|