Printer with OCR function

Looking for All-in-one printer with OCR functionality, or maybe scanner or specialised OCR hardware if it's not too expensive. Anyone currently using one of those and recommendable?

I´m not a friend of all-in-one solutions. They tend to be weaker than single devices.

We´ve had 2 cheap HPs, a medium priced Canon (800 series with separate tanks I think) and a 450 EUR Brother Laser combo.
The summary is easy: You get what you pay for.
The laser was by far the best and most professional device. No problems at all, except for the hole in the wallet. The Canon is good but it has two problems: the paper feed is unprecise, so much that I suspect we have a faulty product, and the ink cartridges are always empty, although it´s one of the most efficient printers according to the tests.
The 50-100 EUR HPs are just trash. Not much more than a few plastic pieces to hold the expensive cartridges. The only good thing about them is the competent service.

So my conclusions are: buy single devices. Laser is by far preferable to ink if you don´t absolutely have to have color.

Originally Posted by Gorath
I´m not a friend of all-in-one solutions. They tend to be weaker than single devices.
…

So my conclusions are: buy single devices. Laser is by far preferable to ink if you don´t absolutely have to have color.

Completely agree - our experience with all-in-ones has been mediocre at best. We are happy with separate scanner and printer now. As for OCR, that is a software thing, but is definitely impacted by how well the scanner does capturing the image.

My inlaws have a Samsung SCX-4100, and they're pretty happy with it. It's a small laser with a scanner, and also functions as a standalone photocopier. Does nothing spectacularly well, but everything reasonably.

I used to have an Epson inkjet all-in-one, and it was just awful in every possible way.

I also recommend dedicated devices unless your needs are very specific.

OCR is one of those applications that sounds great but has never worked worth a damn. Maybe someday.

Is that so? How far is the accuracy of current OCR software in recognizing roman letters, or it's the problem of cheap printers or scanners unable to support OCR to function properly?

The big problem with OCR is the time it takes to clean up after it. Its mistakes aren't in the least bit intuitive, and that makes proof-reading a lot harder. You need a good scanner, expensive software and documents that are in very good shape.

Honestly, I haven't tried the latest stuff so I really shouldn't insist. But I feel justified, anyway. I was first interested in it back in the x486 days and experimented with it extensively. Then I would check it out again every now and then over the years. It always disappointed.

OCR is a great idea that people want to believe in it (I know, because I was one of them).

It depends on how fault tolerant you are, in my experience with a reasonable quality origional running the result through a spell checker will catch most of the errors but something will occassionally slip through.

And it all depends on what you intend to use it for.

For example, suppose you run a small business or own a couple of houses you rent out. You'll need to keep a track of a fair bunch of contracts, receipts, documents, and what not. With an OCR-enabled scanner, you can scan them in for references as bitmaps -- and simultaneously OCR them to make them searchable. Then you can just use Google Desktop or whatever to do full-text searches on them. This can be very, very useful when you need to find a particular document quickly. The typos won't matter, because it'll get enough words right to make the search a working proposition.

(N.b.: this is not a theoretical example. I don't (need to) do this myself, but I know people who do, and it works really well.)

That's the idea, and it sounds great. Everybody likes that idea.

There's only one catch, and it's one everybody always tends to underestimate. That's the time it takes to clean up documents that, for one reason or another, don't scan very well.

I'm speaking from firsthand experience too. I read fast (Evelyn Wood), type fast (about 90 words per minute or so) and know how to use a PC. I should be able to make OCR my bitch (had to say that).

But it even took me forever to clean up OCR documents, sometimes. Faxes, copies of copies, tiny or unusual fonts -- there are some things OCR has a tough time handling.

Oh, and everyone always swears by it. I'm not just being argumentative, either. Everybody always wants to tell you how great OCR is working for them.

All you can do is experiment with it. If it's going to be valuable to you, you'll know it in less than a week.

Wow, that's some serious typing. I can manage about 80 WPM on a good day, and really don't know how much faster I could get. That said, I don't *think* much faster than 80 WPM, so I'm not sure how much added value there would be in it anyway…

You reach a point where you type one thing while thinking another. That may sound hard, but it's not at all. It's like holding two thoughts at the same time. How hard is that, really?

Depends on the person, I guess. Personally, I have something of a one-track mind. Wiser not to chew gum when walking, that sort of thing. You should see me try to dance…

I'd guess that one reason that it's a pain to error-check OCR is that our reading process (particularly fast reading) is fairly fault tolerant, especially to errors that arent in the first or last character. But for the kind of application that PJ describes you'd probably get by without checking the entire document, as your potential search keys would be in a few well defined subsections of the document.

And in any case a few typos won't seriously degrade the searchability -- most of the words will be scanned in correctly, and if you search with more than one term, you'll certainly get good hits.

Anyway, it works quite well in practice, without having to manually spell-check the documents.