Image scan doesn't detect lowercase "L"

[Chmedly 1]

I've recently noticed that Evernote can't find notes containing words that have a type face where the lowercase "L" has the little "flag" at the top but nothing at the bottom.  It seems that it isn't recognizing this character as an "L".   I'm not certain yet if it is thinking it's a "1" or not.  Any one else notice this?  

 

For instance, this receipt shows "grilled cheese".  Evernote can't find it if I type "grilled" but it CAN find it if I type "cheese".  

[jbenson2 2,147]

Lower case L is not the problem.

The accuracy of the character recognition is based on several factors - such as image quality. 

 

Here is a receipt with grilled that Evernote found and highlighted. 

As with the Chmedly's example, the letter i does not have a serif at the bottom.

[Chmedly 1]

You're right.  I spent some more time playing with it and although it seems quite legible to me, that particular note doesn't want to be searched.  The only word that it seems to recognize on the whole receipt is "cheese".  Until now I've been very impressed with how well it can recognize search terms in a scan and I've actually been relying on it for that.   This has me wondering what other documents there are that I can't find.  

 

Is there a way to submit notes like these to the team so the algorithm can get better?  

[jbenson2 2,147]

Occasionally Evernote employees visit this user forum. They might see this.

Or you could contact them via twitter @evernotehelps

 

There are additional ways to improve your results.

I find the search results are best when using a good quality receipt. 

To improve the quality of thermal paper receipts, I usually rely on CamScanner (an app for my phone). It will crop and then enhance the image. I send it to Evernote in a PDF format.

 

I have found that the OCR for PDF's  is better than for JPG's.

As you have found out, OCR is not a black & white science, It is more of an artform with lots of variables.

 

To increase my search result accuracy, I also use very specific tags, typed-in keywords, descriptive titles, and accurate Create Dates,