(Archived) precise searches

[patnpm 30]

If I type in a search term - such as 'manual' - I get a list of all the manuals I've stored in evernote, but also a number of OCRed documents in which there are handwritten words which have been interpreted as approximating to 'manual' - some of them bearing little resemblance to the original term. This gets worse the shorter the search term.

Its not really a problem at the moment but as my database swells I can see it becoming quite frustrating to find the document I want amongst all the guesses of the OCR system. So I wonder if there would be any way of excluding OCRed documents from a search? That way I can get only the printed documents with the search term I want. I know I can already restrict searches to document titles with the 'intitle' search syntax.

This is NOT a complaint about accuracy of the OCR system, which I'm sure must have a degree of laxity in order to make it workable.

[crane 40]

I would love this option too - I end up wasting a lot of time looking through search results, ignoring any terms highlighted in an image.

[engberg 89]

You could eliminate notes that contain images by adding this to your search:

-resource:image/*

Unfortunately, I doubt that's what you want, since that will also reject clipped web pages, etc.

I usually just type another word or two from my target document to eliminate false matches, but there's currently no way to explicitly rule out text within recognition results.

We will be rolling out some new image processing technology on the service in the next few months that should give much higher quality results for good-quality images and scans. (I.e. fewer false positives.)