Jump to content

Search not accurate IOS 10.4


Recommended Posts

  • Level 5*

There was an issue in earlier versions of EN wherein the server OCR engine was not as effective as a desktop engine.  As such search results on IOS platforms were not as complete as those on the desktop.  This issue still exists.  Trouble ticket from 2018 is 2621943.

Also, on my iPhone running 10.4 a search like "S e a r c h" including the parentheses returns notes containing what looks like any of the letters (10,000 notes).  Not what one would expect.  On my iPad running 8.24 the expected no notes is the result.

Link to comment
  • Level 5

My state about OCR is this: If a pdf contains a text level (which means it was created with it, or got it by OCR), the server will not OCR it again. So the OCR should happen either before EN (typically with the software delivered with the scanner), or on the server, but not twice.

This would exclude OCR as a source of search differences.

I am not sure about iOS search: When online it probably uses the server based search. But if offline content is available, it searches in flight mode as well, so it has a local search capacity. And this works better than the server - it does find content in notebooks shared to me by another person (if downloaded as well), which the server based search does not. So if you want better search results on EN iOS v10, get everything downloaded to offline, and switch to flight mode !

I have a ticket running on this.

Link to comment
  • Level 5*
8 hours ago, PinkElephant said:

My state about OCR is this: If a pdf contains a text level (which means it was created with it, or got it by OCR), the server will not OCR it again. So the OCR should happen either before EN (typically with the software delivered with the scanner), or on the server, but not twice.

This would exclude OCR as a source of search differences.

I am not sure about iOS search: When online it probably uses the server based search. But if offline content is available, it searches in flight mode as well, so it has a local search capacity. And this works better than the server - it does find content in notebooks shared to me by another person (if downloaded as well), which the server based search does not. So if you want better search results on EN iOS v10, get everything downloaded to offline, and switch to flight mode !

I have a ticket running on this.

Not sure but I don’t think that’s the case.  The PDF’s at issue were OCR’d on the desktop.  If what you say were true the notes containing those PDF’s would appear in search results on IOS.  They didn’t and still don’t.  Plus got this statement during 2018 trouble ticket.

The issue you're seeing is related to server-side search, which seems to have an issue with recognizing the wrong font when scanning PDF files. The text in the PDF is printed in a monospaced font, so the letters are wider, but the OCR sometimes incorrectly recognizes it as a non-monospace font, so it sees the gap in position and thinks there's a space there.

The point of this post is worst case if the OCR mechanism still has the same issue then server based search results with V10, desktop and mobile, will be incomplete.  At least for PDF’s containing monospace fonts.

 

  • Like 1
Link to comment
  • Level 5

Well, I mainly rely on pre-OCRed pdfs.

The ix500 came with a copy of Abbys FineReader; my ScannerPro on mobile has OCR build in; these two cover my main import of scanned documents.

So hard for me to tell if there are server issues with OCRing new content. But good to know, I will keep an eye on it.

Link to comment
  • Level 5*
9 minutes ago, PinkElephant said:

Well, I mainly rely on pre-OCRed pdfs.

The ix500 came with a copy of Abbys FineReader; my ScannerPro on mobile has OCR build in; these two cover my main import of scanned documents.

So hard for me to tell if there are server issues with OCRing new content. But good to know, I will keep an eye on it.

I also fully rely on "pre-OCRd" PDFs.  Which is why I saw the issue in 2018.  I was searching for something via my phone which I knew existed but could not find it.  Went to the desktop and there it was.  The trouble ticket ensued. 

A little disconcerting that the engine in use at the EN servers is not as robust as it could be.  V10 being a server based search product, mostly.  With a pile of notes for less specific searches you don't realize you have an incomplete search result.  And Murphy being Murphy, the one you need will be the one that's missing.  :wacko: 

  • Like 1
Link to comment
  • Level 5

True, search must be reliable. What I don’t know is how they handle language recognition.

On my OCR clients, setting the language really improves the OCR results. This is even true between European languages, especially of the romantic and northern family. Most OCR programs are supported by dictionary search, and good ones by statistical methods to guess on which word will be close to another. Knowing the language helps a lot with both.

GoodNotes follows another approach when OCRing handwritten text. They create several guesses on a single word, and put them invisibly into the document. So they will find the searched text, because it usually is one of the 5, 6 or more possible words saved in OCR. You are happy it found what you were looking for, and are not aware there were other option just waiting to correct what maybe went wrong in OCR. 

Will keep an eye on it. But since I do my main work with EN still on legacy, it will probably skip my attention.

  • Like 1
Link to comment
  • Level 5*
1 hour ago, PinkElephant said:

True, search must be reliable. What I don’t know is how they handle language recognition.

On my OCR clients, setting the language really improves the OCR results. This is even true between European languages, especially of the romantic and northern family. Most OCR programs are supported by dictionary search, and good ones by statistical methods to guess on which word will be close to another. Knowing the language helps a lot with both.

GoodNotes follows another approach when OCRing handwritten text. They create several guesses on a single word, and put them invisibly into the document. So they will find the searched text, because it usually is one of the 5, 6 or more possible words saved in OCR. You are happy it found what you were looking for, and are not aware there were other option just waiting to correct what maybe went wrong in OCR. 

Will keep an eye on it. But since I do my main work with EN still on legacy, it will probably skip my attention.

For sure.  Though the aforementioned issue seemed to have more to do with how the OCR engine interpreted characters as opposed to words.  IAC, I am sticking with 6.25.1 for now, for some reason I don't like referring to it as legacy???  :(

Link to comment
  • 2 months later...
  • Level 5*

Search not working so well with iPhone IOS 10.7 either.  I was out and about and needed to find something.  I entered the a one word search on my phone and could not find the note from September I knew should be there.  Got to my desktop (6.25.1) and did the same single word search and found the note.  In fact the search yields 369 notes on desktop and 212 on the phone.  I can't check V10 on the web but 5.33 says 212 notes as well. 

Something woefully amiss here.  I think it may be that EN OCR of PDFs does not work well.  I've seen this problem in the past where local OCR is more accurate/complete than EN server OCR.  157 notes is a lot to miss though.  Scary thing is desktop V10 may be missing the same157 notes in search results if it is using the server information.

Link to comment
  • Level 5*

Is this a discussion of search when online or offline?   
My understanding is that if online, all searches are run on the server

IOS or Desktop only applies when offline

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...