SongKong Jaikoz

SongKong and Jaikoz Music Tagger Community Forum

Problem with MusicBrainz and Discogs specified album matching

#1

I have discovered a problem with how Jaikoz matches to specified albums, either Music Brainz or Discogs. It appears to be a problem if the release has duplicate track titles. This is very common with special edition releases were you have bonus material with different mixes or live versions.

I have ripped my media without metadata except track and disc numbers. I then use the “Match Songs to Specified Album” action.

When is has completed I have discovered that the tracks are matched incorrectly.

For the following album:
Release “Young Americans” by David Bowie - MusicBrainz

  • Disc one is matched correctly except for track 2 which is matched to track 13 on disc 2 which has the same name.
  • On disc two only tracks 23-25 are matched correctly. tracks 1-11 are matched to tracks 12-22 on the release and vice-versa.
#2

Hi, could you please run Advanced:Create Support Files and send support@jthink.net the JaikozSupport.zip file

#3

Actually you do seem to have title set but is set wrong e.g as just Chapter01, Chapter02, or Title01

This is a pretty difficult release https://musicbrainz.org/release/c38eb25c-9d85-3a9d-a344-61b57d4b8930 because you basically have three different versions of the same track on the same release, although I cant tell exactly why it got this wrong.

If you would like me to analysis further I can but I would really need you to upload your release using Dropbox so I can test myself and see how I can improve the results.

#4

Thanks for your quick response.

I realise that this issue will only impact a small number of releases. I have recently been ripping my DVD audio releases to add them into Roon and it seems that quick a number of them are anniversary releases which contain the original album plus bonus material, for example live or alternate mixes of the sae tracks. My concern is that there is flaw in the matching logic which which means that the album will be ripped with tracks out of sequence and/or incorrect metadata.

The names of the tracks has no bearing on the end result. I have tried multiple attempts having the tracks named perfectly and it still mixes the tracks up.

I will upload to drop box. and let you know when I have done so.

#5

Thankyou I will try and look later today.

#6

You are right there is a bug !

For single disc and two disc releases we compare every every song with every track and then work out a match that allows every song to be matched to one track and with the best score. (Because the number of possible combinations gets too large we use a different method if release has more than two discs).

But we are comparing song track no with the contiguous track no of the release, the contiguous track no is what the track no would be if we treated the release as a single disc release, so in your example that would be tracks 1-35. This makes no difference for disc 1 but for disc 2 it means that for this release (with 11 tracks on first disc and 25 tracks on second) the first song on disc2 is should match 11th track based on disc no/ track no but instead matches the 1st track and the 11th track on disc 2 should match the 22nd track but matches the 11th track.

Now track no is only one part of the score weighting used to compare each song with each track, but because you essentially have the same song twice on disc 2 with in most cases the same duration and the artist is the same for all tracks there is little to differentiate so in some cases the track no match is causing the wrong mapping to be selected.

What we should match against disc no/track no if songs have disc no/track no or try matching against track no and then contiguous track no if songs only have track no.

An additional problem I noted if if your songs do not have values in the disc no or track no fields then we try and match without this information, but in your case this information can be parsed from the filename so we should try and parse this information.

Raised https://jthink.atlassian.net/browse/JAIKOZ-1410 and should be fixed soon for next release.

#7

Fixed as part of Jaikoz 11.5 Radiohead released May 11th 2022