Mutiple word search, none highlighted. Updated

[doctorkeo 25]

Version 10.17.6

I want to be able to search for multiple words in a text jpg image, and have the found words highlighted like it does for single words – I really thought this was solved, or has it just been ‘broken’ with v10?

Example in the search box I put “West Park”, and Evernote finds several notes with the phrase, but none are highlighted. Also, when I do a ‘find within note’ on the found notes, none are found. A ‘find within note’ using a single word West or Park does confirm the phrase is in the note. What am I doing wrong?

,

Thanks

Doc.

I see I've had a few views of this question, but no answers - Is this a stupid question, or does no one have answer? Any reply is better than none.

[Mike P 2,619]

All I can do is to agree that there seems to be an issue with searching for multiple words within images. Sometimes neither word is highlighted and sometimes one but never both. I think find within document (ctrl-F) will only find text rather than words within images. As you can no longer look at the results of the OCR within the enex file I don't really know how to progress. I suggest a support ticket.

[PinkElephant 8,066]

It finds and highlights in most cases all words when I try. Sometimes it only highlights one, but these are the exceptions.

We should remember there are limits to the picture OCR, both conceptually and because it may have a problem reading a specific word among others.

[doctorkeo 25]

I will raise the 'feature' with the support team, but I can't see how they are NOT aware of the problem.

What I can't understand is how other users who really depend on the EN data like businesses and serious research projects manage.  For me all the data supports hobbies - mostly genealogy. I really hope Evernote get their act together. For me I'm trapped by Evernote, I've tried time and time again to export my data into other programs like OneNote and Nimbus, but I never got more than half the data across - and much of the exported notes are gobbledygook. 

[PinkElephant 8,066]

All OCR is limited, especially in pictures. What is easy for a human to identify may be difficult for an algorithm to „see“. EN has a tremendous hit rate, I think it is way above 99%.

Then there is the way the index is build. Some embed a full text layer „above“ the picture of the text. Others build a list of words and their positions in the document. Then a word will be listed once, followed by a list of positions.

Some apps even build a list of alternative words - for example GoodNotes 5 is doing it when OCRing handwriting. This improves the recognition rate, but can as well lead to false positives (highlighting wrong words). EN has very little false positives, so I think they don’t use this strategy.

My observation is described above. It may be that one or another occurrence is already missed in the OCR. It could be as well that the highlighter sometimes jumps a position (which would be a bug).

Since I use mostly PDFs and usually OCR them outside of EN, testing it is not that easy for me. Let’s wait what support says.