question_pair / README.md
datalama's picture
Update README.md
3ee1279
---
license: mit
dataset_info:
features:
- name: id
dtype: int64
- name: question1
dtype: string
- name: question2
dtype: string
- name: label
dtype:
class_label:
names:
'0': 다른 질문
'1': 같은 질문
splits:
- name: train
num_bytes: 634021
num_examples: 6888
- name: test
num_bytes: 62628
num_examples: 688
download_size: 403049
dataset_size: 696649
---
## Dataset Description
songys님이 오픈소스로 공개한 Question_pair 데이터셋을 약간의 데이터 수정을 거쳐 업로드한 데이터셋.
원본 데이터셋과 자세한 설명은 아래 repo 참고
- **Repository: https://github.com/songys/Question_pair**
**수정 사항**
- `is_duplicate`를 `label`이라는 필드로 rename함.
- test set의 `test_id``id`로 rename함.
- 기존 0, 1에 대한 label을 반대로 변경함.
- as-is
- {"같은 질문": 0, "다른 질문": 1}
- to-be
- {"같은 질문": 1, "다른 질문": 0}
- 최종 field는 'id', 'question1', 'question2', 'label'를 선택하여 저장함.
## Dataset Structure
```
DatasetDict({
train: Dataset({
features: ['id', 'question1', 'question2', 'label'],
num_rows: 6888
})
test: Dataset({
features: ['id', 'question1', 'question2', 'label'],
num_rows: 688
})
})
```