OCR and PDF text extraction

Make a scanned document searchable, or pull the text out of it.

Extract text from a PDFGet the words out as plain textMake a scanned PDF searchableRecognise text in a scanImage to TextCopy text from a pictureImage to WordPhoto of a page to Word

Which one do you need?

How to tell which kind you have

Open the file and try to select a line of text with your mouse. If a neat blue selection follows the words, there is a text layer and extraction will work. If nothing happens, or you select a rectangle covering the whole page as though it were one picture, it is a scan.

The appearance tells you nothing: a scan of a laser-printed page can look cleaner on screen than a genuine digital document. Only the selection test is reliable, and it takes two seconds.

A document can also be partly both — a born-digital report with a scanned appendix stapled onto the end is extremely common, and it extracts perfectly for forty pages and then returns nothing.

The mistake people make

Trusting OCR output without reading it. OCR is a very good guess and it is still a guess. The characters it confuses most are the ones that matter most: 0 and O, 1 and l and I, 5 and S, 8 and B. A misread digit in a reference number or an amount does not look wrong anywhere — the sentence still reads normally — so it survives a proof-read that would catch a misspelled word. Check figures individually against the original.

Worth reading before you start

The three things people most often wish they had known about this kind of job.

All articles in the knowledge base

Related categories