|
|
--- |
|
|
dataset_info: |
|
|
features: |
|
|
- name: image |
|
|
dtype: image |
|
|
- name: text |
|
|
dtype: string |
|
|
- name: markdown |
|
|
dtype: string |
|
|
- name: html |
|
|
dtype: string |
|
|
- name: file_name |
|
|
dtype: string |
|
|
- name: ocr |
|
|
dtype: string |
|
|
splits: |
|
|
- name: train |
|
|
num_bytes: 27893994569.17 |
|
|
num_examples: 113091 |
|
|
download_size: 25517477643 |
|
|
dataset_size: 27893994569.17 |
|
|
configs: |
|
|
- config_name: default |
|
|
data_files: |
|
|
- split: train |
|
|
path: data/train-* |
|
|
--- |
|
|
# HuCCPDF |
|
|
Approximately 113k pages of Hungarian PDFs from the Common Crawl, as featured in our paper "Synthetic Document Question Answering in Hungarian". The `text` field is extracted using PyMuPDF, and the `ocr` field is extracted using pytesseract. |
|
|
|
|
|
See other datasets from the paper: |
|
|
- HuDocVQA (https://huggingface.co/datasets/jlli/HuDocVQA) |
|
|
- HuDocVQA-manual (https://huggingface.co/datasets/jlli/HuDocVQA-manual) |
|
|
|
