Arabic printed document and form OCR data
Printed Arabic and bilingual documents as offices produce them: letters, forms, certificates, reports and tables, scanned, faxed and photographed. Every block is labelled with its role and reading order, every table cell is read, every form field is matched to its value.
Specification
- Documents
- Letters, application forms, certificates, reports, statements and tables
- Layout
- Right-to-left, mixed Arabic–English, multi-column, stamps and signatures
- Images
- Scans, faxes and phone photos, JPEG, PNG or PDF
- Annotation
- Layout blocks and reading order, table structure, form key–value pairs, full transcription
- Privacy
- Personal details replaced before delivery
- Quality check
- Every document is cross-checked by several trained reviewers
Example of the data
- Field
- الاسم الكامل
- Value
- سارة أحمد
- Block
- header · 1st in reading order
- Table
- 4 columns × 6 rows
What it is used for
Document AI
Train layout, table and form understanding for Arabic documents.
Onboarding and KYC paperwork
Read application forms and certificates without retyping.
Records digitisation
Turn paper archives into structured, searchable data.
RAG over documents
Feed language models with documents read in the right order.
Questions
How is reading order handled?
Every block is numbered in the order a native reader follows, right to left and across columns, including English inserts.
Are stamps and signatures labelled?
Yes, as their own block types, so models learn to skip or keep them.
Can I see samples first?
Yes. Request samples with the form on this page and we will send them with the full specification.
Related datasets
Tell us what your model needs.
Request samples and the full specification. We reply within a couple of business days.
Request samples