Dataset Viewer
Auto-converted to Parquet Duplicate
text
large_stringlengths
1.59k
4.71M
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\nGIFT FROM THE LIBRARY OF MAJID FAKHRY\n2009\n\n---\n\n\n\n---\n\(...TRUNCATED)
"A.U.B. LIBRARY\n\nAMERICAN\nUNIVERSITY OF\nBEIRUT\n\nUNIVERSITY\nLIBRARY\n\n---\n\n\n\n---\n\n\n\n-(...TRUNCATED)
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\n3149.2\n\nH341A\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n-(...TRUNCATED)
"309.156\n\nH24wA\n\nJA 26\n\nAG 24\n\nFEB 2\n\n21 Sep 64\n\n31 Aug 67\n\n5 Dec 68\n\n4 NOV\n\nJAFET(...TRUNCATED)
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\nUNIVERSITY\nLIBRARY\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n---\n\n\(...TRUNCATED)
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\nUNIVERSITY\nLIBRARY\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n---\n\n\(...TRUNCATED)
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\nUNIVERSITY\nLIBRARY\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n---\n\n\(...TRUNCATED)
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\nUNIVERSITY\nLIBRARY\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n---\n\n\(...TRUNCATED)
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\nUNIVERSITY\nLIBRARY\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n---\n\n\(...TRUNCATED)
"AMERICAN\nUNIVERSITY OF\nBEIRUT\n\nUNIVERSITY\nLIBRARY\n\n---\n\n\n\n---\n\n\n\n---\n\n\n\n---\n\n\(...TRUNCATED)
End of preview. Expand in Data Studio

nyu-aco-ocr-full

Scanned Arabic books from the Arabic Collections Online (ACO) archive, OCR'd page by page with the dots.ocr vision-language model served through vLLM. The archive spans seven partner collections: NYU, Princeton, Cornell, Columbia, AUB, AUC, and UAE National Archives.

Each PDF page is rendered at 200 DPI, OCR'd individually, and the pages of a book are joined into one markdown document separated by \n\n---\n\n. Each row is one complete book.

Schema

Column Type Description
text string Full book text in markdown, pages joined in order
source string Book identifier

Usage

from datasets import load_dataset

ds = load_dataset("AdaMLLab/nyu-aco-ocr-full", split="train", streaming=True)

A note on the dataset viewer

The text can look scrambled in the dataset viewer and in raw dumps, but inspecting the sample in full shows it is well preserved.

License

The scans come from the ACO archive, which distributes public domain and openly available Arabic books. Refer to Arabic Collections Online for the terms covering individual works.

Downloads last month
173