---
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)