Jump to content

Problem with OCR of handwritten text


Recommended Posts

hi everyone,

i just bought evernote premium, mostly for its OCR feature. however, i can't get OCR of my handwritten notes to work. i realize that i have very bad handwriting, so i tested the OCR with a simple note with only four clearly (for my standards anyway) written words. When searching for each of these words, EN recognizes "death star" just fine, but neither "jupiter" nor "wizard" are recognized. is there a problem with my handwriting or is this a limitation of EN's OCR algorithm?

i have this problem both with EN desktop 6.8.7.6387 on win10 and with EN web. I have already tried to CTRL + help --> fix this note, but it did not change the results.

i have bought EN premium specifically for the OCR feature, so this is a bit disappointing. do you have any ideas how to fix this?

thanks a lot!

dsocr.jpg

  • Like 1
Link to comment
  • Level 5*

Hi.  In this copy it looks like there are three things -

  1. The paper comes out as dark grey against the black writing.  Is there some way to get a plain white background?  That might be better paper,  or more light!
  2. Where text OCR is critical,  lines are a bad thing - at least those visible in the scanned file are.  Pale blue lines may scan far less visibly than darker ones.
  3. Text recognition always is a little hit and miss.  I've saved some pre-Evernote paper notebooks into my notes,  but each section has a typed summary or keywords to help me find the right page.  My eyes are then quicker than repeated (failed) searches to scan through and find the right paragraph.

 

 

Link to comment
  • Level 5*

The OCR process runs on the server, and generates a .reco file for each image5a566f60bdb41_ScreenShot2018-01-10at11_48_01.png.f802f100ba8d731f1858627cca1408b8.png
You can see this file on a Mac, or the data if you export to a .enex file
There's documentation at https://help.evernote.com/hc/en-us/articles/208314518-How-Evernote-makes-text-inside-images-searchable

Note: I believe this is language specific; I've identified my language as English - Your language would be ?

I took a screenshot of your note, and this is what the OCR process guessed at

<!DOCTYPE recoIndex PUBLIC "SYSTEM" "http://xml.evernote.com/pub/recoIndex.dtd"><recoIndex docType="unknown" objType="image" objID="ed9a6bbd4fb2bfefa44a53ad1af05a84" engineVersion="7.0.24.1" recoType="service" lang="en" objWidth="178" objHeight="197">
<item x="33" y="39" w="39" h="16"><t w="59">death</t><t w="55">Dem</t><t w="54">dab</t><t w="52">deal</t><t w="51">demi</t><t w="48">am</t><t w="47">DM</t></item>
<item x="31" y="78" w="62" h="24"><t w="45">zip ? ?</t><t w="44">fig ? ?</t><t w="43">jug ? ?</t><t w="41">grip ? ?</t><t w="37">zip IR</t><t w="37">fig IR</t><t w="37">GNP ? ?</t><t w="37">spir</t><t w="36">jug IR</t><t w="36">grip IR</t></item>
<item x="30" y="115" w="47" h="12"><t w="54">wind</t><t w="49">wizard</t><t w="44">Winn</t><t w="43">wined</t><t w="40">indian</t><t w="39">wins</t><t w="38">union</t><t w="38">hind</t><t w="38">wild</t><t w="13">li.:)'1\0,cl</t></item>
</recoIndex>

  • Like 1
Link to comment

following user gazumped's suggestion, i tried again with a new note and a better photo with higher contrast. EN seems to recognize the text in that note just fine, so for now i just blame my previous problem on the low image quality.

thanks for your help!

  • Like 1
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...