|
--- |
|
language: ko |
|
license: cc-by-sa-4.0 |
|
tags: |
|
- mrc |
|
- korean |
|
- finance |
|
- legal |
|
dataset_info: |
|
- config_name: multiple_choice |
|
features: |
|
- name: doc_title |
|
dtype: string |
|
- name: context |
|
dtype: string |
|
- name: question |
|
dtype: string |
|
- name: options |
|
sequence: string |
|
- name: answer_text |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 52657661 |
|
num_examples: 32092 |
|
- name: test |
|
num_bytes: 8296685 |
|
num_examples: 4634 |
|
download_size: 14848868 |
|
dataset_size: 60954346 |
|
- config_name: span_extraction |
|
features: |
|
- name: doc_title |
|
dtype: string |
|
- name: context |
|
dtype: string |
|
- name: question |
|
dtype: string |
|
- name: answer_text |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 154553286 |
|
num_examples: 93828 |
|
- name: test |
|
num_bytes: 23273377 |
|
num_examples: 12692 |
|
download_size: 43600501 |
|
dataset_size: 177826663 |
|
- config_name: span_extraction_how |
|
features: |
|
- name: doc_title |
|
dtype: string |
|
- name: context |
|
dtype: string |
|
- name: question |
|
dtype: string |
|
- name: answer_text |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 111843943 |
|
num_examples: 61746 |
|
- name: test |
|
num_bytes: 21025535 |
|
num_examples: 9124 |
|
download_size: 31577500 |
|
dataset_size: 132869478 |
|
- config_name: tableqa |
|
features: |
|
- name: doc_title |
|
dtype: string |
|
- name: context |
|
dtype: string |
|
- name: table_title |
|
dtype: string |
|
- name: table_html |
|
dtype: string |
|
- name: question |
|
dtype: string |
|
- name: answer_text |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 228901199 |
|
num_examples: 96000 |
|
- name: test |
|
num_bytes: 24161924 |
|
num_examples: 12000 |
|
download_size: 15321234 |
|
dataset_size: 253063123 |
|
- config_name: text_entailment |
|
features: |
|
- name: doc_title |
|
dtype: string |
|
- name: context |
|
dtype: string |
|
- name: question |
|
dtype: string |
|
- name: answer_text |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 52307558 |
|
num_examples: 32750 |
|
- name: test |
|
num_bytes: 5778697 |
|
num_examples: 3740 |
|
download_size: 13330652 |
|
dataset_size: 58086255 |
|
configs: |
|
- config_name: multiple_choice |
|
data_files: |
|
- split: train |
|
path: multiple_choice/train-* |
|
- split: test |
|
path: multiple_choice/test-* |
|
- config_name: span_extraction |
|
data_files: |
|
- split: train |
|
path: span_extraction/train-* |
|
- split: test |
|
path: span_extraction/test-* |
|
- config_name: span_extraction_how |
|
data_files: |
|
- split: train |
|
path: span_extraction_how/train-* |
|
- split: test |
|
path: span_extraction_how/test-* |
|
- config_name: tableqa |
|
data_files: |
|
- split: train |
|
path: tableqa/train-* |
|
- split: test |
|
path: tableqa/test-* |
|
- config_name: text_entailment |
|
data_files: |
|
- split: train |
|
path: text_entailment/train-* |
|
- split: test |
|
path: text_entailment/test-* |
|
--- |
|
|
|
# 금융, 법률 문서 기계독해 데이터셋 |
|
|
|
금융 및 법률 분야 전문문서를 활용하여 기계독해 모델 생성을 위한 지문-질문-답변으로 구성된 데이터셋입니다. |
|
|
|
## 데이터셋 구성 |
|
|
|
### Subsets |
|
- **span_extraction**: 정답경계 추출형(정답이 지문 내 특정 범위에 있는 데이터) |
|
- **span_extraction_how**: 절차(방법)(어떻게로 시작하는 질문에 대한 정답 추출 데이터) |
|
- **multiple_choice**: 다지선다형(객관식 형태의 데이터) |
|
- **tableqa**: Table 정답추출형(표 기반 질의응답 데이터) |
|
- **text_entailment**: Yes/No 단문형(Boolean 선택 형태의 데이터) |
|
|
|
### Splits |
|
- **train**: 학습용 데이터셋 |
|
- **validation**: 검증용 데이터셋 |
|
|
|
## 사용 방법 |
|
``` python |
|
from datasets import load_dataset |
|
특정 configuration의 특정 split 로드 |
|
train_data = load_dataset("shchoice/finance-legal-mrc", "span_extraction", split="train") |
|
valid_data = load_dataset("shchoice/finance-legal-mrc", "multiple_choice", split="test") |
|
특정 configuration의 전체 split 로드 |
|
full_dataset = load_dataset("shchoice/finance-legal-mrc", "span_extraction") |
|
``` |
|
|
|
## 데이터 필드 |
|
1. 기계독해에 필요한 필수 항목 |
|
- doc_title: 문서 제목 (Optional) |
|
- context: 문장의 주요 텍스트 내용 |
|
- question: 질문 텍스트 |
|
- answer_text: 질문에 대한 정답 텍스트 |
|
2. 정답 힌트 |
|
- clue_text: 정답의 단서를 제공하는 텍스트 |
|
3. 테이블 QA에 필요한 정보 |
|
- table_title: 표의 제목 |
|
- table_html: 표의 HTML 데이터 |
|
4. 테이블 QA에 필요한 정보 + 정답 힌트 |
|
- cell_text: 정답이 포함된 셀의 텍스트 데이터 |
|
- cell_coordinates: 정답이 포함된 셀의 좌표 데이터 |
|
5. 다지선택형 QA에서 필요한 정보 |
|
- options : 객관식 내용들(1. 정답 2. 오답 3. 오답 4.오답) |
|
6. 문서의 분류명 |
|
- class: 문서의 분류 정보 |
|
- code: 문서의 코드 정보 |
|
|
|
## 라이선스 |
|
이 데이터셋은 CC-BY-SA-4.0 라이선스 하에 제공됩니다. |
|
|