Jump to content
sjpschmid

bug in search

Recommended Posts

Evernote often has incorrect results to searches done over all one's notes.  Here is a simple example.  I searched on the term "aslia", and it returned one note.  But then when you search within that note, it shows there are actually no matches.  The search within a single note seems to always work. 

I have attached two screenshots - one showing the global search result, and the other showing the conflicting search result within the one note.

I have also attached an activity log.

evernote bug 1.PNG

evernote bug 2.PNG

Evernote_6.18.4.8489_20190909.log

Share this post


Link to post

Any images in the note?  Full URL title?

Share this post


Link to post

Any posts on the forum that are related to EN search, get my attention, as I would love to rely on and use EN search more than my current 5% of the time.

So, I took a look at the page/URL you added, and also clipped it, and all but one of the images has any text captured on the image, and that one image, contains the word 'Asilomar'.

So if you search the Page URL for the partial-word "asila' or ""aslia' it too returns no results...same as for your EN Note!  IF you use "asilo" in your Note Search or via 'All Notes', it will indeed return just this one note.

I think your error was a typo in your search query...so this time I reluctantly think the winner is EN

Share this post


Link to post
2 hours ago, sjpschmid said:

Yes, there are images in the note. 

No guarantee but ENs OCR may have found asila in the image.

Share this post


Link to post
4 hours ago, RavBoy said:

Any posts on the forum that are related to EN search, get my attention, as I would love to rely on and use EN search more than my current 5% of the time.

So, I took a look at the page/URL you added, and also clipped it, and all but one of the images has any text captured on the image, and that one image, contains the word 'Asilomar'.

So if you search the Page URL for the partial-word "asila' or ""aslia' it too returns no results...same as for your EN Note!  IF you use "asilo" in your Note Search or via 'All Notes', it will indeed return just this one note.

I think your error was a typo in your search query...so this time I reluctantly think the winner is EN

No, if you look at the screenshot I provided, there was no typo.  I searched for "asila".  And this was just one example.  I have gotten incorrect search results many times.

Share this post


Link to post

So you're saying, that you should be able to find something, right?

When I searched the URL you provided, and also the clipped note I made, and searched (CTRL+F) for 'asila', it returned zero results.  Because there's nothing on the URL-page that has 'asila' in it.  If you change the last character in the search query asila, the 'a' to an 'o' as in 'asilo' for 'Asilomar' then you will find 3 results straight away, hence why i thought you made a typo.  What do you believe is on the page, that should be found when entering 'asila'???

EDIT*

and it was the same searching 'asila' in All Notes - nothing was returned...which i think is your issue, that EN is returning a result that it shouldn't

Share this post


Link to post

Search on OCRed content can never be absolutely precise.

OCR is not done by „reading“ character by character, and putting it together. It will read the dots, build a table of dots => characters and then based on the selected language(s) make a guess on the word itself. If not sure, it will put alternatives into the metadata containing the OCR results.

When searching, EN will find these, which may result in false-positives (notes found that do not contain the word) as well as in false-negatives (notes containing the word, but not found). This is independent from whether the OCR was done before importing into EN, or on the EN servers themselves.

The more languages you choose, the wilder will be the guessed OCR-results, and the further off can be results. On the other side, if a word is not contained in a dictionary (like names, words from a foreign language etc.), OCR guesses will likely be off, finding words by trying a best match to the selected language.

  • Like 2

Share this post


Link to post

Also found no results for aslia after clipping that page to Evernote, both article and full page.

I have run into a similar issue where web copied data contains hidden html text, sometimes I use an html editor to find the match.  In a couple of instances the issue was important enough that I deleted all hidden matches.   Now I generally paste as plain text whenever possible, to avoid this issue.

I suppose it is better to find hidden content, than failing to find content I know is there, though I learned to get around that issue as well.

  • Like 1

Share this post


Link to post
16 hours ago, RavBoy said:

So you're saying, that you should be able to find something, right?

When I searched the URL you provided, and also the clipped note I made, and searched (CTRL+F) for 'asila', it returned zero results.  Because there's nothing on the URL-page that has 'asila' in it.  If you change the last character in the search query asila, the 'a' to an 'o' as in 'asilo' for 'Asilomar' then you will find 3 results straight away, hence why i thought you made a typo.  What do you believe is on the page, that should be found when entering 'asila'???

EDIT*

and it was the same searching 'asila' in All Notes - nothing was returned...which i think is your issue, that EN is returning a result that it shouldn't

Yes, my issue was the "false positive" when searching All Notes.  The correct result was obtained when just searching within the note.  I have experienced this often - where the results of a single note search differ from that of the All Notes search.

Share this post


Link to post
4 hours ago, Don Dz said:

Also found no results for aslia after clipping that page to Evernote, both article and full page.

I have run into a similar issue where web copied data contains hidden html text, sometimes I use an html editor to find the match.  In a couple of instances the issue was important enough that I deleted all hidden matches.   Now I generally paste as plain text whenever possible, to avoid this issue.

I suppose it is better to find hidden content, than failing to find content I know is there, though I learned to get around that issue as well.

Thanks for checking into this.  I agree with your sentiments about being better to find hidden content.  It is strange, though, that it would be found when searching All Notes, and not within just a single note.  I have experienced this behavior many times.

Share this post


Link to post
13 hours ago, PinkElephant said:

Search on OCRed content can never be absolutely precise.

OCR is not done by „reading“ character by character, and putting it together. It will read the dots, build a table of dots => characters and then based on the selected language(s) make a guess on the word itself. If not sure, it will put alternatives into the metadata containing the OCR results.

When searching, EN will find these, which may result in false-positives (notes found that do not contain the word) as well as in false-negatives (notes containing the word, but not found). This is independent from whether the OCR was done before importing into EN, or on the EN servers themselves.

The more languages you choose, the wilder will be the guessed OCR-results, and the further off can be results. On the other side, if a word is not contained in a dictionary (like names, words from a foreign language etc.), OCR guesses will likely be off, finding words by trying a best match to the selected language.

Thanks for your reply.  Are you sure that this content is OCR'd?  I did not think that it was.  Even if it was, it does not seem right that EN gives different results when searching All Notes vs. when searching just within one note.  This has happened to me many times.

Share this post


Link to post

I just made some tests:

When I search on the website itself, the term „aslia“ is NOT found. I loaded the page into Safari on my iPad Pro.

OK, they do not like it when you open the page from an European location. Did those of you not in the EU know that many American companies block requests from the EU since we got ourself a tough data protection legislation ? I really wonder why ...

But I deviate: So VPN on active, beamed myself to the US, next try. All content now legible, but no match. Not with the search function of the page itself, and not with the search of my iPad on that page.

So about what are we discussing here ?  A „false positive“ search result in EN, which means the search found something that is not there ?!

For me, this is much less relevant than „False negatives“, which means the content is there, but not found when searched.

Next I have imported the article into my EN. First as website, and then as pdf version of the website. When I search „aslia“, I get ZERO hits, which means the term is not found in either version. This matches my expectation. When I search „Wappo“, I get one hit in each of the 2 copies. So at least from what I have copied for testing purposes, I only get valid search results, negative and positive.

Share this post


Link to post

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...