How to keep an address's line breaks when you OCR

An address keeps its line breaks in Read by Zaweba without you having to ask. In its Paragraphs and columns mode, blocks of short lines such as addresses keep their breaks while ordinary prose is joined into paragraphs. If you want every line kept exactly as printed, switch to the Single column, line by line mode instead.

It's a small thing that matters a lot. An address squashed onto one line is a nuisance to paste into a form or an envelope template, and a list of names joined into a sentence is worse. Many OCR tools either join everything or join nothing, so you end up fixing one part of the page or the other.

How Read tells an address from a paragraph

When Read reads in Paragraphs and columns mode, it looks at the shape of the lines. Long lines that run to the edge of a column are prose that has wrapped, so Read joins them, rejoins any word broken by a hyphen at the end of a line, and leaves a blank line between paragraphs. A block of short lines, the kind you see in an address, a signature block or a short list, is left with its line breaks in place.

That's why a letter usually comes out well in the default mode: the sender's address and the date at the top stay as separate lines, and the body underneath becomes proper paragraphs.

Going through it

Open the letter, envelope or form in Read: drop it on the window, paste it, scan it or open a PDF. Let the clean-up do its job if it's a photo; the Document tab says what was done.

Leave the layout on Paragraphs and columns for a mix of address and prose. Look at the address in the side-by-side review and you'll see each line on its own line of text. Click on a word in the picture to find it in the text if the page is busy.

If the page is all lines that matter, switch to Single column, line by line. This keeps the page's own line breaks in reading order and doesn't join anything. It suits address labels, lists of names and numbers, delivery notes, and anything else where the line is the unit.

Getting the address out cleanly

Once the text is right, select the address in the text panel and copy it, or copy all the text and trim it where it's going. You can edit the text directly in Read before you copy, and edits are used when you save as well. When Tesseract did the reading, words under 70 percent confidence are shaded amber; postcodes and house numbers are worth a second look anyway, since a misread digit is easy to miss.

For a pile of envelopes or forms, the Batch tab reads a whole folder in the background and saves the results in the formats you've ticked.

Other output formats

A .docx keeps line breaks within paragraphs as real line breaks, so an address stays tidy in Word or in Write by Zaweba. A plain .txt file keeps them too. A searchable PDF keeps the original page exactly as it looks, with invisible text underneath, so copying from it later gives you the words.

Handwritten addresses are another matter. Handwriting is one of the honest limits on Read's Help tab, along with low resolution pictures and decorative fonts, so a hand-addressed envelope may need more correcting. Printed labels and typed letters are what Read by Zaweba handles well.