Server 4.8 and Auto Organize

[kingy444 117]

Emby Server Version: 4.8.0.80
Emby.AutoOrganize, Version=1.7.0.0

Just updated to 4.8 and a file i had sitting in the organise folder was not identified correctly.

I've turned 'Overwrite existing episodes' off as a result but as the file was not identified correctly in v4.7 I assume there have been some 'improvements' to the identification code, are these explained anywere? I am concerned about incorrect identifications resulting in mismatches

Bit concerning as the mismatch took a 1 word title and matched to the a 3 word title. I dont think this is a close enough match to 'assume' you have it right ?

There doesnt appear to be a way for me to 'correct' this manually either. There is no option to add a smart match manually or identify the right series under the match.

How do i tell AO it got it wrong?
And ultimately, I guess - Please help restore my faith in AO getting it right, because it has up until server version 4.8

Edited February 3, 2024 by kingy444

[Luke 42077]

Hi there, can you please provide a specific example?

How to Report a Problem

Thanks.

[Luke 42077]

Quote

How do i tell AO it got it wrong?

Have you explored clicking the organize button?

[kingy444 117]

 

2 hours ago, Luke said:

Hi there, can you please provide a specific example?

How to Report a Problem

Thanks.

The series that is misidentified also exists in the library - The detection of the particular episode isnt of concern, that part makes sense but the series name match is too broad

 

edit: just reiterating that the file was already sitting there as 4.7 had failed to identify - i hadnt gotten around to manually sorting it yet. nothing changed with files or library, just the 4.8 upgrade (and accompanying AO update)

embyserver-AO.zip

Edited February 3, 2024 by kingy444

[kingy444 117]

20 minutes ago, Luke said:

Have you explored clicking the organize button?

No sure how this will help - All the Organize button does is ask if you want toi proceed moving file from X to Y

It doesnt let you override the identification of that episode - Nor am i sure how that would really work (As the file has been sorted at this stage)

The button MAY let you choose if EMBY doesnt know what it is, but because EMBY thinks it knows what it is, i cant override this

I think the only way to ensure it works is to be able to add a manual 'Smart Match' which you dont appear to be able to do

[Luke 42077]

10 hours ago, kingy444 said:

 

The series that is misidentified also exists in the library - The detection of the particular episode isnt of concern, that part makes sense but the series name match is too broad

 

edit: just reiterating that the file was already sitting there as 4.7 had failed to identify - i hadnt gotten around to manually sorting it yet. nothing changed with files or library, just the 4.8 upgrade (and accompanying AO update)

embyserver-AO.zip 21.65 kB · 1 download

What is the example that we're looking at?

[kingy444 117]

4 hours ago, Luke said:

What is the example that we're looking at?

I thought that is why you wanted the log. There is only one attempt to sort in the log. Thanks

[Luke 42077]

4 minutes ago, kingy444 said:

I thought that is why you wanted the log. There is only one attempt to sort in the log. Thanks

Right but it just helps save time if you can spell out that information as well. Thanks.

[kingy444 117]

The phrase "Monsters" Incorrectly translated to the series "Aaah, Real Monsters"

I have also confirmed that when i hit "Organize" on a show that isnt identified, I can correctly identify the series

So to clarify the two issues:

1. Unable to re-identify / correct the series once EMBY has made the decision
2. The matching of a 1 word phrase to a 3 word title

Problem 1 could probably be sorted by adding in the ability to manually add a smart match

For problem 2 - I would think the only way to address such a situation is to have the user confirm a match if the 'score' of the match is below a threshold. in this case only 33% of the title matched. 
4.7 did not make this same judgement. So I assume the core naming in 4.8 is a bit more lenient

The only reason I havent lost data here is because of a difference in filetypes (mkv vs avi) so the file wasnt actually overridden

Edited February 3, 2024 by kingy444

[Luke 42077]

So you already had both series in your library?

[kingy444 117]

The series folder existed and had been identified however thetvdb for the series appears to have been deleted. So makes sense the null match there.

Im not super worried that the identification wasn’t automatic against that series, that happens from time to time.

I would much prefer correcting these manually then getting an incorrect match like this.

My concern being that a single word was matched against 3.
I do not believe it would not be uncommon for other shows to fall to the same mismatch.

 

[Luke 42077]

If you use the server search feature to search for monsters, which one comes up higher in the search results?

[kingy444 117]

Aaah Real Monsters

[kingy444 117]

Further Naming tests - this one is replicated with multiple series and looks introduced in 4.8 aswell

"Series Title" works
"Series  Title" (double spaced) works
"Series-Title" works
"Series.Title" doesnt work

It appears that when a file has a dot in the series title the identification doesnt work. The above was the extent of my test but is easily identifiable.

I have used the "." in the file name for years without issue

[Luke 42077]

It uses the server search feature to find existing items in the library.

[kingy444 117]

So the internal search needs to be improved to treat the “.” The same as “-“ noting that using the “-“ as the delimiter does return series ?

[Luke 42077]

Correct.

[kingy444 117]

I also just checked every single other non alphanumeric character on my keyboard

The “.” Is the only character that appears to have this issue

 

Perhaps an easy fix is to do a .replace(“.”,” “) when doing the search from auto organise ?
 

I assume that sort of change is quicker to make (to the plug-in) than in the internal engine?

 

[Luke 42077]

Quote

Perhaps an easy fix is to do a .replace(“.”,” “) when doing the search from auto organise ?

No, because what if the series has a dot? And what about other areas besides auto-organize. So it has to be looked at more broadly.

[kingy444 117]

4 minutes ago, Luke said:

No, because what if the series has a dot? And what about other areas besides auto-organize. So it has to be looked at more broadly.

I’ve just tested this with a couple series that do in fact have a “-“ in the name and do in fact have a “.” In the name - I think it have found the issue

 

All the other characters appear to treat their replace as a space and shows matching both “series-title” and “series title” are returned

 

When I search for a series with a “.” the results differ. I get the results for “Series.Title” but I also get results for “SeriesTitle” (no space) so it appears there is already a replace type search happening just the “.” character is removed rather than being replacing with a space

[kingy444 117]

Popping back in with another thought because this mismatch could be pretty signifigant. Another mismatch I just had made me think of this, this time saved from file loss by metadata not being available for the mismatched target series

My analysis:
Previously in Auto Organise with Emby 4.7 the matching was pretty accurate because we were searching the provider metadata directly by the filename and not querying the local library

Because Emby 4.8 is now using the internal library it is going to be somewhat difficult to ensure you have an exact match unless it is the only result returned, even then it may be wrong because you may be wanting to add a secondary series with a similar name (and the "Create new series" goes out the window in that case).

From my understanding of the issue here we simply cannot assume the first result is the correct one

For instance a couple examples based on the TVDB names. I am sure this problem extends way past the below examples.
One Piece vs One Piece (2023) vs The One Piece
The Office (US) vs The Office vs The Office (PL)
The Avengers vs The Avengers: Earths Mightiest Heroes vs The Avengers: United They Stand
Charmed vs Charmed (2018) vs Charmed to Death 
Lost vs Lost! vs Lost (2021)
 

[AaronClarke 3]

I can confirm I am also experiencing the same issue. TV shows which do not have '.' (Full stop) in the filename are being organised automatically but if there is a full stop in place of a space, they fail. Replacing the full stop with an underscore in the filename allowed the process to complete. See screenshot, Halo completed with no intervention but Chicago PD , Chicago Med and Chicago Fire all needed to be altered manually.

 

[AaronClarke 3]

Further to my previous... this evening I am unable to add Mr. and Mrs. Smith (2024) as there are Full Stops in the actual title of the show. Removing them from the file name does not work when the actual show title includes Full Stops.

[Luke 42077]

2 minutes ago, AaronClarke said:

Further to my previous... this evening I am unable to add Mr. and Mrs. Smith (2024) as there are Full Stops in the actual title of the show. Removing them from the file name does not work when the actual show title includes Full Stops.

Hi, server log?

[AaronClarke 3]

Sorry Luke... was just fetching it... took a while to upload

embyserver.txt