Arabic newspaper OCR: columns read in the right order
Newspaper pages are hard: several columns, headlines across them, captions and ads. LexaOCR reads each article in the order a person would.
The problem
- General OCR tools often read straight across the page, mixing lines from neighbouring columns.
- Headlines, bylines, captions and adverts interrupt the text.
- Small print and cheap paper make dots and diacritics fragile.
How LexaOCR handles it
- The layout agent finds columns and gutters and orders them right to left, top to bottom.
- Two printed lines that touch are separated before reading, so lines are never merged.
- Every line is read by several readers and voted word by word.
- Headings, paragraphs and captions are kept as separate blocks in the Word file.
What you get
- Articles as running text in reading order
- Headlines as Word headings
- A review edition for checking against the page
- Searchable PDF of the original
Questions
Does it handle mixed Arabic and English pages?
Yes. Each line's language and direction are detected separately, so English lines stay left to right inside an Arabic page.
Can I process a whole archive?
Yes. Upload PDFs of up to 1 GB each. For large archives, contact us about the Institution plan.
Try it on one page
The quality check is free, and nothing is read until you confirm.