Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Project Naptha: a browser extension that enables text selection on any image

projectnaptha.com

Post reply on HN
Print view View original

101–110 of 141 posts

Re: Project Naptha: a browser extension that enables text selection on any image

#101

Post by ishi »

username42 wrote: post #98

Just tested with a random scanned page ( http://www.hpl.hp.com/research/info_theory/ShannonWeb/fullsi... ) the result is almost garbage. It seems as bad as most OCR software I have encountered. This was to be expected as it is based on ocrad.

Almost garbage? This is the OCR result for the 2nd paragraph. Almost perfect, although the last word in each line gets joined to the first one in the next line:

"The fundamental problem of communication is that of reproducing atone point either exactly or approximately a message selected at anotherpoint. Frequently the messages have meamlng; that is they refer to or arecorrelated according to some system with certain physical or conceptualentities. These semantic aspects of communication are irrelevant to theengineering problem. The significant aspect is that the actual message isone selected from a set of possible messages. The system must be designedto operate for each possible selection, not just the one which will actuallybe chosen since this is unknown at the time of design."

Re: Project Naptha: a browser extension that enables text selection on any image

#102

Post by entropy_ »

ishi wrote: post #101
username42 wrote: post #98

Just tested with a random scanned page ( http://www.hpl.hp.com/research/info_theory/ShannonWeb/fullsi... ) the result is almost garbage. It seems as bad as most OCR software I have encountered. This was to be expected as it is based on ocrad.

Almost garbage? This is the OCR result for the 2nd paragraph. Almost perfect, although the last word in each line gets joined to the first one in the next line: "The fundamental problem of communication is that of reproducing atone point either exactly or approximately a message selected at anotherpoint. Frequently the messages have meamlng; that is they refer to or arecorrelated according to some system with certain…

I tried it with both ocrad and tesseract modes, and indeed, the ocrad mode produces garbage, the tessaract mode produces a really good result but takes a longer time doing it(mainly the time it takes to upload the entire thing and get the result back).

That seems to make sense to me, at least. Use ocrad mode by default, if it doesn't perform well, switch to tessaract and you'll hopefully get a better result.

Re: Project Naptha: a browser extension that enables text selection on any image

#104

Post by bigbugbag »

username42 wrote: post #98

Just tested with a random scanned page ( http://www.hpl.hp.com/research/info_theory/ShannonWeb/fullsi... ) the result is almost garbage. It seems as bad as most OCR software I have encountered. This was to be expected as it is based on ocrad.

Thanks! I wanted to try not sure it would fare better than usual OCR but was denied as I'm not a google product.

Re: Project Naptha: a browser extension that enables text selection on any image

#105

Post by netcan »

WalterBright wrote: post #70
vidarh wrote: post #61

Earlier quoted context omitted.

There's actually been a bit of research on the error rates you need to beat for OCR to be cost-effective vs. having people re-type. I don't have the references handy, but I believe it's generally cost effective to OCR with error rates up to nearly 2%, and most current "consumer grade" OCR is well below 1% error rates for scans that aren't absolutely atrociously poor quality. My Msc thesis was on reducing OCR error ra…

It was suggested to me by a friend that to get good OCR results, run it through the scanner/OCR twice, then diff the results. Usually one or the other will get it right, and if you run the two results through a difference editor like 'meld', it's quick to fix.

If this is done all in software (IE, it isn't analyzing a slightly different image), why wouldn't the OCR just do this itself?

Re: Project Naptha: a browser extension that enables text selection on any image

#108

Post by erikrothoff »

elwell wrote: post #62

Every time I click "Allow" on "Access data on all sites" for an extension I creep closer to my security hole paranoia threshold. If it was all in JS, who cares? But this sends ajax to remote servers of course. Am I alone?

That is the wording that Google Chrome chose for "allow this extension to access the DOM on any page". It sounds bad but these are the permissions an extension needs to be able to access images and text on any page.

Re: Project Naptha: a browser extension that enables text selection on any image

#109

Post by username42 »

ishi wrote: post #101
username42 wrote: post #98

Just tested with a random scanned page ( http://www.hpl.hp.com/research/info_theory/ShannonWeb/fullsi... ) the result is almost garbage. It seems as bad as most OCR software I have encountered. This was to be expected as it is based on ocrad.

Almost garbage? This is the OCR result for the 2nd paragraph. Almost perfect, although the last word in each line gets joined to the first one in the next line: "The fundamental problem of communication is that of reproducing atone point either exactly or approximately a message selected at anotherpoint. Frequently the messages have meamlng; that is they refer to or arecorrelated according to some system with certain…

When I did the test, it was garbage. Since your answer, I have repeated my test with results similar to yours.

Re: Project Naptha: a browser extension that enables text selection on any image

#110

Post by antimatter15 »

elwell wrote: post #62

Every time I click "Allow" on "Access data on all sites" for an extension I creep closer to my security hole paranoia threshold. If it was all in JS, who cares? But this sends ajax to remote servers of course. Am I alone?

Checking the "Disable Lookup" item on the settings menu prevents it making ajax calls to any server and does all processing locally. Of course there's a resulting drop in speed and OCR accuracy. The lookup requests are all HTTPS, are never logged, and contain no user identifying information.
Post reply on HN