I recenlty volunteered to key in a 20 page school-policy manual for which the original file was lost. (Floppy got corrupted! Remember those?) I was going to key it in, when curiosity and “programmer’s laziness” (a.k.a., “don’t do it manually if it is faster to automate it”) got me thinking about optical character recognition (OCR).
So I downloaded a trial version of SimpleOCR. I scanned the original printed pages and ran the OCR program on it. It is pretty neat. It splits the page into two panes: original text in one, deciphered text in the other. In the deciphered text, it highlights words it is not “too sure” about and words it did not find in its dictionary. You simply go along and correct what it got wrong. It made the job much faster.
The program is far from perfect. Every output line had a few errors. Yet, I was impressed, because one could see that just a little additional work could improve it substantially. One such example: if it deciphered 5 instances of “sveotshirt”, and I change the first to "“sweatshirt”, it ought to default the next 4 to “sweatshirt” as well.
It has a handwriting-recognition mode. My handwriting! I doubt I’ll even try.
I recenlty volunteered to key in a 20 page school-policy manual for which the original file was lost. (Floppy got corrupted! Remember those?) I was going to key it in, when curiosity and “programmer’s laziness” (a.k.a., “don’t do it manually if it is faster to automate it”) got me thinking about optical character recognition (OCR).
The couple of best OCR programs around now, to my knowledge, are Finereader, and Omnipage. Typereader is an excellent OCR program as well, but getting very long in the tooth, last updated in 2000. Any one of them would probably give better results than SimpleOCR, by your description.
Thanks for the info on those products. Next time I have a need for OCR, I will surely check them out.
I just hit upon SimpleOCR, tried it and found it was good enough for my little 20-page task. (I think I typed “free OCR” into Google :dough: . ) If I had to do something like your Objectivism CD, SimpleOCR would have been too error-prone.
Thanks for the info on those products. Next time I have a need for OCR, I will surely check them out.
I just hit upon SimpleOCR, tried it and found it was good enough for my little 20-page task. (I think I typed “free OCR” into Google . ) If I had to do something like your Objectivism CD, SimpleOCR would have been too error-prone.
Actually, I used an earlier version of Typereader for the CD-ROM.
My fiance is studying for the bar and her preferred method is studying from flash cards. Of course, all her material is in long paper form so we tried to OCR a ton of material in order to condense it and put it on cards. SimpleOCR was ok to good but the one we really had the most success with was OmnipagePro.
It was exceedingly pricey but it worked incredibly well and helped especially with some of the notes she took in the margin or on parts of the typed page that were highlighted. Most of the other OCR programs seemed to freak out with highlighter and underlined or circled text but OmnipagePro just passed them right by.
Still, we ended up settling on the most optimal solution of all: we hired a friend that was an unemployed writer to transcribe the notes for us. It saved time and alot of loading pages into the scanner etc. Sometimes the optimal solution is simplest one.
In fact, I still use them quite alot, as it about the most convenient way to move around files from home to the university computers…until I get a USB Thumb Drive in the next few days. It might just be the floppies I buy, but the current floppies seem more fragile these days.
Who remembers those horrible 6 inch floppies? Now, they were nasty, big and yet they could only hold 640K.
And, to be on topic for a moment: Yes, it does sound like a really good program. Sure, far from perfect apparently, but which of these things is near perfect? I might download it and give it a try, and even see if it likes my sometimes messy handwriting. Ill let you know how it turns out if I remember to.
This BBC story (HT: BoingBoing) is about a device that combines OCR and text-to-voice technology to create a camera-like hand-held device that can be pointed at text, which it will then read aloud. Useful for blind people who want to read. [i remember reading that Stevie Wonder owned a machine of this type. It might be the large $50,000 type mentioned in the article.] This new version is smaller, and cheaper – at $5,000. Pretty cool.
Meanwhile, voice-to-text seems to improving too, if this review is to be believed.
And then, there is this BMW system that uses thermal imaging to help drivers spot objects at night.
If you have Microsoft Office 2003, there is excellent OCR support in there. It’s part of the MS Document Imaging tools.
We use it through the COM library where I work and it does great on printed text. The only complaint I have is that it has horrible problems with handwriting. If there is a signature on a paper it sometimes will freeze during the OCR process.
But if you are only scanning text, it works great.