Datasets:
Tasks:
Text Classification
Formats:
parquet
Sub-tasks:
sentiment-classification
Languages:
Korean
Size:
100K - 1M
License:
Update README.md
Browse files
README.md
CHANGED
@@ -4,13 +4,21 @@ Korean sentiment classification dataset
|
|
4 |
- Size: 460K(+180K)
|
5 |
- Language: Korean-centric
|
6 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
### ⚠️ Caution with `Rating` Column
|
8 |
0 stands for negative review, 1 stands for positive review, and 2 stands for ambiguous review.
|
9 |
**Note that rating 2 is not intended to be used directly for supervised learning(classification).** This data is included for additional pre-training purpose or other usage.
|
10 |
-
In other words, this dataset is basically a **binary** sentiment
|
11 |
|
12 |
### 🔍 See More
|
13 |
-
See all the codes for
|
14 |
See Kaggle dataset in [Kaggle Dataset](https://www.kaggle.com/ninetyninenewton/kr3-korean-restaurant-reviews-with-ratings).
|
15 |
|
16 |
### Legal Issues
|
|
|
4 |
- Size: 460K(+180K)
|
5 |
- Language: Korean-centric
|
6 |
|
7 |
+
### Usage
|
8 |
+
```python
|
9 |
+
from datasets import load_dataset
|
10 |
+
|
11 |
+
kr3 = load_dataset("Wittgensteinian/KR3", name='kr3', split='train')
|
12 |
+
```
|
13 |
+
|
14 |
+
|
15 |
### ⚠️ Caution with `Rating` Column
|
16 |
0 stands for negative review, 1 stands for positive review, and 2 stands for ambiguous review.
|
17 |
**Note that rating 2 is not intended to be used directly for supervised learning(classification).** This data is included for additional pre-training purpose or other usage.
|
18 |
+
In other words, this dataset is basically a **binary** sentiment classification task where labels are 0 and 1.
|
19 |
|
20 |
### 🔍 See More
|
21 |
+
See all the codes for crawling/preprocessing the dataset and experiments with KR3 in [Gitlab Repo](https://gitlab.com/Wittgensteinian/kr3).
|
22 |
See Kaggle dataset in [Kaggle Dataset](https://www.kaggle.com/ninetyninenewton/kr3-korean-restaurant-reviews-with-ratings).
|
23 |
|
24 |
### Legal Issues
|