As impressive as Evernote’s ability to read handwriting (even my awful scrawl, occasionally) is, there was always something about it that didn’t seem quite right. Firstly, that it could recognise writing in situations that nothing else can. Also, it seemed odd that it would convert your writing to text, but not give you access to that text afterwards.
Evernote themselves only ever said your text was made searchable – they never said they’d convert it to text.
I’d suspected for a while what was really going on. I figured it might be working out what the images could match, but not being too specific – knowing that some images could match various different combinations of letters is probably a lot easier than working out which ones they actually match.
I wasn’t sure, though, until I searched on two completely different words, and both highlighted the same bit of text. I’d written ‘Tuesday’, and scanned the page. It matched for a search for ‘Tuesday’ perfectly. It also highlighted the word ‘Tuesday’ when I searched for ‘Testing’.
Evernote has no idea what the text says, but it can still make it searchable. They’ve been very clever, and realised that they don’t need to actually read handwriting in order to make it searchable. They just have to match to anything that might say what you’re searching for, on the grounds that false positives will be fairly rare, and don’t actually matter.
The technology behind it isn’t as amazing as it first seems, but the thinking that went behind the technology is brilliantly simple.