Text Grab - $ symbol and columns

  • 1
  • Idea
  • Updated 1 year ago
  • (Edited)
The Text Grab function is awesome. After upgrading I've already used it 3 times this morning.

Refinements possible:
1. One area it needs improving is interrupting the '$' symbol. I took a screenshot of my bank statement to paste into an excel spreadsheet. The $ was replaced with a 5 or S. So I zoomed in my browser and did it again, this time all the $ were changed to S, better. When I super zoomed (massive text), all $ were read as $.

2. On the first occasion when I grabbed 3 numbers in a column they then pasted into 3 excel columns (awesome!). On the next 8 number screenshot grab, the 8 numbers would only paste into 2 line. This working well would be a real asset. I was so impressed it did do it with the 3 numbers.
Photo of Paul

Paul

  • 14 Posts
  • 6 Reply Likes
  • excited with the new feature

Posted 1 year ago

  • 1
Photo of george

george

  • 86 Posts
  • 12 Reply Likes

It could well be the font used.  Looking at the text of your message, the Dollar sign ($) just looks like the letter S with a dot over it.

But if I zoom IE to 160%, I can see it OK.  I suspect that if you tried using Arial Unicode MS font. where the Dollar has a full vertical line, you would have no problem.

Photo of Paul

Paul

  • 14 Posts
  • 6 Reply Likes
In the example I gave I don't have choice of what font the bank displays its bank statements with. The '$' font we see here will be set by the css webpages chosen by TechSmith. It's a good example though, of how hard the $ sign could be for the OCR to recognise.
Photo of george

george

  • 86 Posts
  • 12 Reply Likes
Believe me, working in the blindness field, I'm very familiar with OCR problems.  A feature which some commercial programs have is the option to define a specific character.  By default, a character which is consistently wrong, can be highlighted, and the software told that if it sees this "glyph", apply a specific character - in this case Unicode 0024 (Dollar sign).
Photo of Daniel Foster

Daniel Foster, Snagit strategy lead

  • 861 Posts
  • 192 Reply Likes
Thanks for the detailed feedback, Paul. I know the dev team is continuing to look at refinements so specifics on how the OCR is performing or where it's struggling for users is helpful.

Generally speaking, I have noticed that 10 or 12 pt text is recognized more accurately than 8 pt text. The amount of contrast between text and background can also make a difference. For one project I needed to grab text from a PDF and was able to set the zoom a little larger in my PDF viewer, which gave me noticeably better OCR results.