Skip to content

Dataset · Delores-Lin

Multilingual document parsing benchmark

Delores-Lin/MDPBench

The first benchmark for evaluating recognition and parsing of real digital and photographed documents across 6 languages.

What's inside

Contains digital and photographed documents with text in six languages: Chinese, English, Arabic, German, Spanish, and French. The dataset was created as a systematic test to check models in real rather than sterile conditions with different writing systems.

What you can do with it

  • Evaluating the quality of multimodal models and OCR systems on complex scans and photos
  • Comparing document parsing accuracy across different languages and scripts
Tasks
Image → text
Languages
Chinese, English, Arabic, German, Spanish, French, Hindi, Indonesian and 4 more
Downloads this month
6.3K
Likes
60

How to load it

from datasets import load_dataset

ds = load_dataset("Delores-Lin/MDPBench")