Tables and forms OCR for Arabic and Urdu documents
Scanned tables come back as real tables, and forms come back as labelled fields, in the right-to-left order of the original.
The problem
- Most OCR returns tables as loose lines of text that have to be rebuilt by hand.
- Right-to-left tables often come back with their columns reversed.
How LexaOCR handles it
- The formatter agent recognises tables, forms and invoices from the page image.
- Cells are filled from the verified words, so every number is checked like any other word.
- Tables are written right to left in Word, matching the original.
What you get
- Word tables with the original rows and columns
- Form fields as label and value pairs
- JSON with every word and its position, for developers
Questions
Can I get the data into Excel?
Copy the Word table into Excel, or use the JSON export with every word and its position.
Try it on one page
The quality check is free, and nothing is read until you confirm.