SARD: A Large-Scale Synthetic Arabic OCR Dataset for Book-Style Text Recognition
Yasser Alhabashi, Omer Nacar, Serry Sibaee, Adel Ammar, Wadii Boulila · 2026
Read SARD: A Large-Scale Synthetic Arabic OCR Dataset for Book-Style Text Recognition by Yasser Alhabashi, Omer Nacar, Serry Sibaee, Adel Ammar, Wadii Boulila (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13588801.txt.
Description
Yasser Alhabashi and his four co-authors, publishing in 2026, release a giant training set for teaching computers to read Arabic books. Optical character recognition for Arabic lags behind other scripts. Existing datasets cover single words or lines, not full book pages, and lack variety in typefaces. The team built their dataset synthetically instead, generating clean page images by computer.
The result, called SARD, holds over 2.6 million document images and nearly 795 million words. Six fonts are represented, from modern screen faces to traditional calligraphic styles. Because the pages are rendered, not scanned, they carry no stains or skew, and the set can grow on demand. The authors benchmark several recognition systems on it, old and new. They offer SARD as shared ground for building Arabic readers that can handle real book layouts.
Catalog record
| CREATOR | Yasser Alhabashi, Omer Nacar, Serry Sibaee, Adel Ammar, Wadii Boulila |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |