How to OCR two-column text in the right order
Read by Zaweba's Paragraphs and columns mode reads one column after another, so newsletters and articles come out in order, and stay as columns in .docx.
To read two-column text in the right order, use Read by Zaweba with its Paragraphs and columns layout mode. Read finds the columns, reads the first one top to bottom and then the next, and joins the wrapped lines back into paragraphs. When you save as .docx, the columns are kept as columns.
Plenty of simple OCR tools read straight across the page. On a newsletter, a journal article or a church magazine, that produces lines that hop from the left column to the right and back again, half a sentence from each. Untangling it by hand takes longer than typing it out.
What it takes
Read by Zaweba on a Windows PC and the page: a scan, a phone photo, a picture or a scanned PDF. If it's a photo of a printed page, Read's clean-up will find the page, flatten it and even out shadows before reading, which helps column detection a lot because columns need straight edges to show up.
The method
Open the page in Read. For anything with columns, the Auto engine setting chooses Tesseract, the bundled engine, because it does layout analysis. You'll see the engine and the reason for choosing it above the text. If Windows OCR was used for some reason, choose Read again with Tesseract.
Make sure the layout mode is Paragraphs and columns. In this mode Read joins lines into paragraphs, rejoins words that were split with a hyphen at the end of a line, puts a blank line between paragraphs and reads columns one after another. The text you get reads like the article did, rather than like a pile of short lines.
Check the result in the side-by-side review. Click a word in the picture and it's selected in the text, and the other way round, which makes it easy to confirm that the second column starts where you expect. Uncertain words are shaded amber; F8 jumps to the next one.
Saving it
How you save depends on what's next. Copy the text or save a .txt if you just want the words in order, say to quote from. Save a .docx if you want something that looks like the page: Read writes two-column sections into the document, along with paragraphs, tables and page breaks, without needing Word. From there, Open in Write by Zaweba hands the document straight to Write, or to whatever program opens .docx files on your PC.
A searchable PDF is the other choice. The page looks exactly as scanned, with the words laid underneath as invisible text, so you can search it and copy from it.
Where it can go wrong
Column detection relies on a clear gap between columns. Very tight gutters, pictures that cross both columns, or pull quotes dropped into the middle can confuse the order. Have a look at the start of the second column in the review and fix anything out of place before you save.
Some layouts are better read differently. A form with labels on the left and answers on the right isn't really two columns of prose; there, Single column, line by line keeps the page's own lines in reading order and may give you something more useful. And if the page is really a table, Read can spot rows of words lined up across wide gaps and save them as a table in .docx or as CSV.
A few pages or a lot
One article at a time is fine in the main window. For a run of back issues, put the scans in a folder and use the Batch tab, which works through them in the background with the same layout mode and saves the results in the formats you've ticked. It all happens on your PC, offline, with Read by Zaweba.
