File size: 888 Bytes
8680c61 219014b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 |
---
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)
|