OCR can read the main language on a page correctly but get a second alphabet, a name written in another alphabet, or a two-language form label wrong. Check each language separately and test sample text so you can find and fix these mistakes.

Step by step

  1. Find where each language is.See whether the languages change by page, column, paragraph, or field. A two-language form may need different checks from a letter written in two languages.
  2. Choose the right languages.Turn on the languages that appear on the page, and do not add a long list of other languages, which gives OCR more ways to misread words.
  3. Check sample text.Test accents, non-Latin alphabets, names, numbers, and mixed punctuation. Open the image beside the recognized text.
  4. Know what the exported text is good for.Searchable text helps you find pages, but keep the page image as the record when the reading order, or the text in one of the languages, may be wrong.

Check before you send

Search for a known word from each language and copy a sample into a text field. If OCR keeps getting one part of the page wrong, add a note on the page or make a separate crop of that part, and do not assume all the text is equally accurate.

  • You know where each language appears.
  • All enabled languages appear on the page.
  • Sample text in each alphabet was checked.
  • The page image is kept.
In Paperlane

OCR in several languages helps you find documents faster. It does not replace a person checking names, legal wording, and translations.

How to review mixed-language pages

A bilingual form may use one language for labels and another for names, addresses, or instructions. Check each part of the page separately instead of assuming the main language is used in every field.

Test one sample from each alphabet, including accents, mixed punctuation, and numbers beside words. If a page switches languages by column, check reading order as well as character accuracy.

How to decide

Retake the photo, or check that part separately, when OCR keeps getting one part of the page wrong. The searchable text can still help you find pages while you check the second language by hand or with a separate crop.

Try this first

Keep the page image so everyone checks against the same thing. It shows which label, in each language, belongs to which value.

Example

Imagine a bilingual form uses English labels, French instructions, and a name with accents. If OCR is set to the main language only, it may find some words but get the important fields wrong without any warning.

Check the part of the page in each language and keep the original layout. The searchable text helps you find pages, but do not assume it is equally accurate for every language.

Questions people ask

Should I split languages into separate PDFs?

Only when the recipient or your process needs it. Keeping the original document in one piece keeps the languages next to each other, as on the page.

Why is one script accurate and another poor?

OCR supports some alphabets better than others, and font, resolution, and layout also differ. Improve the photo of the part OCR reads worst and check it separately.

Should I split a multilingual document into separate PDFs?

Only when the recipient needs separate files. Keeping the original order can be more useful when the languages explain each other on the same page.

Use recognized text to find documents, not as a replacement for the page. Compare names, numbers and layout with the scan before you reuse or share the text.