SongKong Jaikoz

SongKong and Jaikoz Music Tagger Community Forum

Problems with "The Messiah"

Attempting to tag “The Messiah” and running into several issues. Songkong manages to identify Part I correctly, but neither of parts II or III, and does not set “Work” correctly in the latter two cases. It also fails to obtain “Movement no” for number 50. I don’t see anything wrong with the MB entry, which does have the correct “part of” relationships for recordings belonging to Parts II and III.

Could I ask Paul to take a look and help identify what is going wrong here?

Can you run Create Support Files please.

Do you have then now?

Yes, thankyou but unlikely to look at it today

Your folder strcuture is not helping and preventing Match to One Album fixing both CDs in one go

Test/George Frideric Handel - Messiah-CD1
Test/George Frideric Handel - Messiah-CD2

can you change so that CDs have shared parent and retry Match to One Album

e.g

/Test/George Frideric Handel - Messiah
Test/George Frideric Handel - Messiah/CD1
Test/George Frideric Handel - Messiah/CD2

Thanks for retrying with master folder, its not clear to me why it can not add that info. You’ll have to upload the files and email support@jthink.net so i can test it out locally, thanks.

Okay so it is quite complicated

So we prefer to derive the Work from the Recording Title, this is because if the work is in the title this is likely to be how you know the work, whereas the connected Works may look rather different, the obvious difference being that the title will probably be in users language whereas the connected Works will be in the composers language (although interestingly dont seem to have this problem for the most famous Russian Composers)

For example this is a Work by Bach with title 6 kleine Präludien: Präludium E-Dur, BWV 937
https://musicbrainz.org/work/671aeca6-05b4-3a40-823b-0d12b2a2a308

And listed below you can see how the title appears for various recordings of the work such as Prelude in E major, BWV 937 or Six Little Preludes

So, to reiterate we try to derive work from title, if we cannot we use MB Works. We know the MB Recording Work linked to the recording (the Movement) and we know its parents work (the Work) so we know a particular group of tracks are all part of the same work, if we cannot derive the same work from the title for all tracks in the group consistently we use the parent work of the MB Recording Work linked to the Recording.

Now in this case the titles actually show the overall work Messiah, HWV 56 rather than the three parts that make up the overall Work and so deriving work from title gives us Messiah, HWV 56.

But for tracks 1-19 we could not derive work from title for two tracks

  • Messiah, HWV 56: I. Sinfony (Grave: Allegro Moderato)
  • Messiah, HWV 56: IX. Air (Contralto) and Chorus: “O Thou That Tellest Good Tidings”

because they have a second colon, so cannot reliably work out work just by looking at this title, so because we cannot find a consistent work for all tracks on title we use the linked MB Recording Work parents work, and for the first nineteen tracks this is Messiah, HWV 56: Part I

We dont have the colon issue for the tracks making up Part II and Part III so we just use the work name derived from title which is Messiah, HWV 56

Possible Solutions
If we look at the MusicBrainz Work tab of the spreadsheet we can see that all tracks have the same value for MB Work, and the right part for the MB Work Level 1

So maybe your solution is use Scripter task to get works from there when not happy wih the one derived from title.

Possibly we could add option to always get Work from MB Work Level 1 Work but then we have the issue of not using Work names that match title.

Maybe better would just be to ensure consistency within release, so in this case since we we could only derive work from title from parts 2 and parts 3 I think we should probably use the MusicBrainz work method for all tracks?

You could also say that we should set the Work to Messiah, HWV 56 since that is how all tracks are titled, and the roman labelling is done from as contiguous numbers. And this is indeed what would have happened if it wasn’t for those extra colons.

Now extra parsing for handling the extra colons is just not viaible there will always be cases where the extra colons are relevent/not relevant to a work. But what I thought was in this example we found a consistent work title for 17 of the 19 tracks in part I, and for the two we could not find title if we did a match to find the work name for these two tracks that matched the value found in the other 17 then we would find a consistent match, and therefore would be confident we had found a match and had a consistent work for all tracks in the grouping, so I think I will add this logic.

I agree that the algorithm for extracting Works from Titles is to be used, then it should seek consistency. Finding a common prefix (terminated by a colon?) between adjacent tracks would, I imagine be successful almost all of the time. It should of course handle the case where two or more separate works are present on a release, and I think the criterion of adjacency would help minimise mistaken identifications.

It would also solve the associated issue of getting the track title wrong by inadvertently breaking at the wrong parse point.

As a possible point for further consideration, I would consider a “first colon in string” as the break point between work and track title would make more sense than a “last colon”. Though no doubt we could find a counter-example if we looked.

Another example where there are two colons is
https://musicbrainz.org/release/441cf12c-064e-4b43-9afe-ef3c46381138
In this case again the overall work is consistent before the first colon, the part "Kyrie, Gloria, Credo, Sanctus and Agnus Dei) between the first and second colons and the individual title following. The formatting of the Roman track number before the part might not be so easy to handle.

Yes, that is what I do

In this case we would not try to derive Work Name from title, because this is effectively free text it is a fools errand trying to handle something like this, instead we would just make use of MusicBrainz Works.

I’ve raised two issues and they are fixed for next release:

https://jthink.atlassian.net/browse/SONGKONG-2887
https://jthink.atlassian.net/browse/SONGKONG-2888

Thank you. Do you have a plan for the release schedule?

Probably within a few weeks.