The only performer listed for the album in the main panel is Andrew Davis. Interestingly the first person mentioned in the review is George Szell, the conductor of Mahler’s fourth symphony, the main feature on the CD.
The way information is presented on the album screen differs with each album. Sometimes its the conductor only, sometimes it is the orchestra, sometimes the composer, then there are combinations in all possible variations.
To me it would seem to be desireable that the presentation is consistent, for example:
The same haphazard way of presenting information can be seen further down in the list of compositions. Performer seems to always contain the orchestra, sometimes the soloist, sometimes not.
Here also it would seem desirable to list contributors in a consistent way.
The point that really gets me is that in the credits (per album, but also per composition or per track) the information is all there. It simply needs to be presented in consistent and easy to read way.
Would it be possible to make these changes to the UI for classical music?
I suspect the root of the issue is poor metadata from the source used by Roon. This seems to be a reissue of an earlier recording. There’s also a remastered version of Szell’s recording that is available without the Andrew Davis addition that Sony appear to have simply tacked on. This has somewhat more accurate metadata, however, the review is missing. @joel - room for improvement here?
Thanks, @Geoff_Coupe. It doesn’t really look like faulty metadata to me, as the credits are all there, incl. Szell, Orchestra etc. The credits even list soloists per composition / track. It would seem that little effort would be needed to extract the relevant information from the credits and list it on the screen in a structured way.
Separate to my point about how information is presented, I agree with you that the metadata provided by Rovi on allmusic.com is neither complete or always correct. @joel I have hinted politely several times in this forum that a structured dialogue between Roon and Rovi might be helpful, which makes Rovi aware of the shortcomings and ensures that the metadata is continually improved. @joel I am not at all suggesting that Roon start correcting metadata, but there needs to be a structured process to continually improve the source data.
That is probably the case.
However my request was not directly concerning improvements to the metadata itself (that might be the subject of another request). I was suggesting that the data contained in the metadata / the credits be displayed in a more consistent way.
The Stables effect must be quite pronounced. In another thread four months ago I made Roon aware that the metadata of Malcolm Sargent’s Pictures at an Exhibition is wrong:
This was @joel’s response at the time:
No change since then. I have even changed the track names of the album to correspond exactly to the allmusic templates, even numbering the different Promenades, so they can be kept apart. Pictures still gets divided into three parts. Interestingly the Tidal and Qobuz versions of the exact same album list Pictures correctly in one piece. This would indicate that the metadata Roon have access to is not wrong.
I don’t want to be impolite, but I have my doubts about the structured process working.
There must be something fundamentally wrong here.
Interestingly I observe the following:
If I re-identify the tracks of Pictures (after having named the track title as it is on allmusic), Roon very briefly display the CORRECT composition, which is
Pictures at an Exhibition (Kartinki s vïstavski), for orchestra, orchestrated by Ravel
Before reverting to the wrong composition, which is cut in three parts.
I am fully aware that reconciling the metadata that comes from Qobuz, Tidal and Rovi / allmusic is no easy task. But something seems to be going wrong here.
And I can well imagine what it is that is going wrong here:
The Malcolm Sargent Pictures / Nights on Bald Mountain is NOT on allmusic,com (an extended 2 CD version is and it has the correct metadata). Roon has metadata for this album, but it is faulty. When I ask roon to identify the compositions on this album, it finds the compositions on allmusic and briefly displays them. Then it finds its own (faulty) metadata and displays that.
The only way to fix this problem is to
remove the faulty (old) metadata
get Rovi to list the album on allmusic.com with the correct metadata
If there is indeed a structured process in place, this should be easily fixable in four months.
I have given up hope for “better” metadata by external suppliers long ago. Therefore I am heavily tagging manually. It’s not always fun, but it also leads to new discoveries…
I think what we should hope for in Roon is a better structure and an easier and more transparent way to map existing local data to the Roon DB structure. Composition tags like Instrumentation, Form and Period come to mind.
Currently it is very often a hit and miss due to all the Roon magic happening.
The metadata for classical on allmusic is actually not all that bad. They are being quite consistent in identifying and naming compositions. I found that if an album is not in allmusic and I enter the composition and movement names as track name, the composition is usually identified correctly and can be cross referenced across my library.
Its outliers like the Malcolm Sargent Pictures that do not work this way, because there is something off in Roon’s database (the one they use besides allmusic).
That’s why I am so adamant that Roon should implement a structured process feeding back errors to Rovi / allmusic and having them fix the issues at the source. That’s the only way this will ever get fixed.
Should be easy and also in Rovi’s interest. They claim metadata is their core business.
The other thing that would be needed is an easy way to extract composition data from allmusic.com and copy it over to albums not in allmusic.com using jRiver et. al. The manual process is way too complicated.
I am not arguing with that. However, in order to achieve this I had to retag my collection (> 200K Tracks) with the composition names used by allmusic. I have still 15% of unidentified albums - which admittedly is a big improvement over when I started using Roon 3.5 years ago.
But there is no real consistency towards composition tags. Just look at the different threads on how to handle single opus collections like the Chopin Nocturnes or Beethoven Quartets. There are often different ways of grouping those, dependent on the album releases.
Form a composition database I expect consistency. There are a lot of situations where I needed to “override” allmusic tags with mine in order to achieve this.
Since there is currently no way in Roon to group those types of compositions together, I am tagging my compositions on the “lowest level” and keep custom grouping tags for op.xx collections. It’s a herculean task to bring consistency of any kind into the metadata and I am realistic enough to not expect much effort will be put into classical. At the end classical is not the top priority for most of metadata providers anyway.
I do agree thatt his needs to be tackled at the source, but what kind of fixes do you expect to happen? Just correcting obvious errors (like having a wrong composition assigned)? As soon as we are talking consistency you’ll need a kind of rulebook and stick to it…
Look at what allmusic has for the Chopin Etudes
Is op.10 and op.25 the correct composition or is it op.10/9 and the other individual pieces? These are not real errors, but they make identifying and presenting information in Roon more difficult unless someone defines a hierarchy and sticks to it. Some albums have the grouped op.10, others use individual compositions.
So why would Rovi do this? To please the classical music Roonies? I’m skeptical. Happy to be proven wrong, though
Finally, classical metadata is much more than just composition tags. There are many more challenges ahaed, like operas and role assignment to performers, and additional composition information like year, form, instruments and period (which would be much more useful than genres).
I hope Roon will find a way to offer a useable mix of improving metadata content and logic as well as giving the user more flexibility towards filling the Roon database with file tag content.
The opera thing is real. There is no way to differentiate between complete operas and selections of discrete arias, overtures etc. Manually I have a SubGenre set as Opera , Opera excerpts so that I can split off complete operas. Opera is a good example as it’s common to see split off pieces , eg overtures. Also it may be good to have a tag for Act then subordinate Parts
Roon metadata on the whole is reasonable but it’s the odd inconsistencies that spoil the experience like Klaus” example of the Chopin Etudes, the Nocturnes and Mazurkas have the same issues despite correcting each part to the Allmusic format and then renaming the Name Tag to this standardized format, so Composition = Etudes Op.10, Part= Etude no.1 in xx, op.10/1, CT.yy etc
It’s a labour of love , a couple of hours to fix Mazurkas for example
The likes of SongKong and MusiCHI tagged help but do not have the same format as AllMusic
By the time you have manually tagged it sorta defeats the object a bit
If only these things could be corrected at the source. Given a suitable interface to the database, I am sure we would all be prepared to help fix the errors.
But that would require a curated feedback process to Rovi / allmusic, which I doubt Rovi will be prepared to implement.
I have albums that have been Unidentified since I started using Roon a few years ago. Am I to believe that they will, in fact, be identified one day? I have lost hope, but I do recall my happy, naive self believing that “there’s a good chance it’ll get a good match in a few days”. I would much rather select @Geoff_Coupe’s album identification… he’s Roon-smart, good looking and erudite…
Then there are albums that are correctly identified, but the composition refuses to be identified, no matter what I do with the tags. Please, if I (or another user) has identified the composition, allow me to use this knowledge…
I still believe that a combination of Roon sourced and crowdsourced data would work. Perhaps I select the Roon metadata to identify the album, but when I see the track listing and/or compositions are messed up, I apply my own (or your previously saved) overlay.
Roon has settled on Rovi / allmusic.com as their primary source of metadata. While this is not perfect, it goes a long way.
That being the case, the only way the quality of that metadata will ever improve is if a structured feedback process is put in place that weeds out errors that users point to.
It is really a pity that this (in my view not unreasonable) request has gone unanswered for a long time. Brians comments are all correct, but general reflections will not improve the metadata. A structured process will.
I don’t think anyone is expecting this stuff to be perfect. But is it really unreasonable to expect obvious errors to be fixed? Roon is a paid service, not a piece of free- or shareware!
Just a last comment. Roon seems to have time to worry about how artists’ photos are cropped, why no time for classical metadata?
You keep saying this. However, (a) there is already a feedback process in place - we report metadata errors, Roon Labs (@vova or @joel) report these to their metadata providers and they get fixed in time. This looks very much like a structured process to me?
And (b) there will always be errors, even with a structured process…