Unicode Support

Many of my classical albums with Russian composers are displaying Cyrillic characters in Roon (i.e., Stravinsky, Tischenko, etc.) To my knowledge, this albums were tagged in dbpoweramp (unicode support).

Will there be a fix for this?

Roon fully supports unicode in file tags.

What I’m thinking is, maybe your album was identified by Roon’s metadata services, and hit an album with Russian metadata.

Can you post a screenshot of the Album page? That should help us get a clearer picture of what happened here.

Example of Cyrillic

Brian, I have quite a few like this. Note the problem after the album title.

Bruce

We have an internal ticket open to track this.

As I suspected before, this is a data problem–one of our metadata providers is giving us data that contains cyrillic, and your album is matching that provider’s data. We have data for this album from another source that does not have a cyrillic artist name–you may be able to find it by clicking the edit pencil and then “identify album” and choosing manually.

In an ideal world, the system would notice that these two versions of the album data are equivalent and automatically merge them together, choosing the data from the more trusted source, which would give you a better result.

This album is difficult because the track titles between the two versions of source data for this album do not match closely, so the automatic system was unable to determine that they were identical.

Brian,

Is there a way to manually correct this?

For classical music, I often get better hits (especially for International Release titles) from Grace Note. For example, I am currently tagging the Pierre Monteux box set on RCA/Sony, and I get no hits from All Music Guide (I use dbpoweramp). However, when I go to gracenote, I get matches for each disc in the 40 CD box set.

Any chance you could incorporate Gracenote?

Bruce

Click the edit pencil on the album screen (in the lower-right corner of the dark section). Click “Identify Album”. Go through the flow–you might find metadata that you agree with in there. If not, you can keep clicking “None of these look right” and then “Use Basic File Information”, which will display information purely based on your file tags.

Over time, we will get better at this stuff…the data problems associated with classical are uniquely challenging compared to more popular content, but we’re committed to making this into a real solution.

Once again, AMG is limited for classical, especially the large box sets Sony, Warner and Universal are reissuing.

You need a better alternative.

I agree.

Gracenote lacks some of the structure (album->track->performance->work->composer) that underlies our data model, and have historically refused to sell us database dumps that we can incorporate into our backend databases, since their offering is API-based. They are no silver bullet here.

Data sources need to be judged on three scales: richness, accuracy, and coverage. Rovi/AllMusic scores high for richness/accuracy and medium for coverage. Discogs is high accuracy/coverage, medium richness. Gracenote is high accuracy/coverage, low richness. MusicBrainz is probably medium/medium/medium and so on.

Incorporating Gracenote data would get us more hits, but that only goes so far, since what we really need are hits with rich data–otherwise we’re not really adding value above what’s already in your file tags. This means proper performance/work/composer structure in the metadata, which Gracenote (and most other providers) do not take seriously.

One idea that I’ve been kicking around for addressing classical/box set situations is to add a second round of identification. After we identify a classical album, if we don’t get work/performance structure for it, we would attempt a second round of identification at the work level that attempts to link up the tracks with work level metadata.

The neat thing about this approach is–while Rovi doesn’t have great coverage of classical box sets, they do have a nice rich list of every piece of western classical music, with descriptions, links to composer data, composition dates, and so forth.

This would enable us to make good use of identifications from other sources, without compromising on well-structured data. It would also normalize our portrayal of the historically significant composers and their works to Rovi’s model, which would be a huge benefit to the browsing experience (currently, you’re seeing a mixture of Rovi’s stuff combined with work structure that was munged within Roon based on your track titles, so there’s strangeness and inconsistency).

One of the upcoming projects in our pipeline is to do a major iteration on the classical browsing experience. It will probably include the work-identification stuff that I mentioned above, along with several more improvements. It’s going to be a little while before we reach that point in the schedule, but it’s on our radar.

Brian,

This is a nice explanation and an intriguing plan going forward.

I have tagged alot of classical cds. Probably the best tool for tagging classical is Musichi Suite. You can get really detailed with this, like opus and all the artists on the recording.

Since there are relatively few new classical recordings on the major labels, most of their new issues are reissues. And to complicate this, most of these reissues come from the European branches.

So, you need to incorporate the breadth of Gracenote with the richness of data from aggregators like AMG. I agree that many of the individual albums comprising these box sets and available on AMG.

I look forward to your progress with this.

Bruce