Questions on Metadata quality...not impressed at the moment at all

I’ll let others decide if your 18 posts in this thread alone constitute a “sermon” or not…I only offer my limited comments, so that the reader can see that not everyone has had your “experience”

And I will simply state again that your comments above are NOT representative of what many here have expressed about the quality of the Metadata and its subsequent presentation within Roon

@Machine: Maybe it’s time to lower your voice a bit. While you do make some valid and well though-out points, both your eclectic tastes and special needs are rather extreme edge cases. It’s still early days for Roon and plenty of ‘mainstream’ challenges are to be met first.

Asking people to ‘stop preaching their sermons’ for gently pointing out to others that you may be the exception instead of the rule is really not helping this discussion go forward. As @mike said, your points have been read and taken under consideration.

It will take time and require some patience, perhaps made bearable by some soothing music, brought to you by Roon.

1 Like

@RBM : Did you read one of my last posts?
“We’ll see what Roon makes out of the informations in this thread, a few
posts before they told me they need some time for figuring out different
issues and we should give them the time…”

What is eclectic about if someone wants to have the exact releases in his library? Some people are grooming metadata for years to get that. Or what do you mean with eclectic?
Why are AC/DC and Pink Floyd - just as an example - edge cases? I am always open for normal discussions.

@Ronnie : Sorry but I was a little bit upset over what you said. Calmed down. I would appreciate a comment for instance if it is enough to you if an album is just identified. I don’t know but for many people this might be the case…

@Machine,

I’m not going to characterize your needs as wacky or eclectic. You clearly take your metadata very seriously, and you’re exactly the kind of customer we would like to serve well. You do have some more specific needs that we aren’t serving well yet. I can’t promise that this will be fixed overnight–we are a small company with pots on the stove, but we do take this stuff seriously, and we appreciate your input.

I think many of the people who are satisfied with Roon’s handling of metadata as it stands today care more about richness than accuracy, and are willing to trade off some accuracy in return for an automatic system that gives a rich experience with little manual labor.

Maybe for some people it’s enough that an album is just identified, maybe also for you. For me it is not!! Most of my albums were identified somehow because there is any release out there from any date that fits to the tracklist… Is it my release? No? Do I want my release in my library? Yes…

Up until now, we have prioritized returning rich, high quality, extended metadata over matching the exact edition that you have, as a conscious product choice. We believe that for most people, hitting high-resolution artwork, a review, detailed credits information, and work-performance-part structure (something that, to my knowledge, no music players other than Roon are attempting to provide today) is more important than returning a more specific, but metadata-poor, exact edition match.

For most popular albums, there are several editions that could conceivably produce identical files and metadata tags in the real world. How do we know whether you have a Japanese release or a US release if they have the same release year, same track titles, same album title, and same track lengths (which in general can’t be trusted due to variations in ripping software)? In the vast majority of real world music collections, the information required to make these distinctions is barely present.

Sometimes we’ll get some information like [Japan] stashed somewhere, or a catalog number, or something. Could be in a directory name, or in a tag, in a sidecar .txt file from a torrent download, or in some other place based on a convention established by an individual person tagging their files. Just as often, there are multiple Japanese releases anyways. After looking at and analyzing millions of real-world files, we have learned: there is no consistent place where we can find reliable information about the original release that a media file came from.

I have ideas about some lightweight things that we could do to improve edition matching specificity and will be discussing them with the team.

Do you have a timeline for the metadata editing already? Will we be able to get custom tags that can be displayed?

It will take us some time to figure out if and when we want to become a platform on which people groom their metadata (as they might have done with various tag editing packages in the past).

Our data model is an order of magnitude larger than any other music player, and building a comprehensive editor for it is a herculean task with an immense amount of UI surface area. I think that it will eventually make sense to do that, but it’s not something we are going to hastily rush out the door.

In the nearer term we are focused on providing solutions for people with highly groomed collections so they can continue to leverage the work they’ve already put into their tags. This is a complex problem–often, tag-based metadata is frustratingly inexpressive, and blending it with a well-structured and precise data model is a difficult engineering problem.

Take a look at our last months of release notes. It should be obvious that we are moving very, very quickly. Releasing some wins on metadata handling issues is one of our next big priorities. I know better than to commit to timelines publicly before work is basically done, so I’m not going to do it. Work is being put into this issue on a daily basis, however. We will get there.

Thanks for your interest and for taking the time to write down your ideas. It has given me some insight into problems in our system that I might not otherwise have.

3 Likes

@brian : Many thanks to your great response to this.

I highly appreciate this.

This is not the point. No one, or for sure me not, is expecting that things happen overnight. I know you are putting a lot of effort into pushing Roon. You can read it here, you can see it in the app.

This is something I tried to explain to @magnuslsjoberg yesterday, it would be a pretty demanding task which can be broke down at least to the matrix code on discs for instance. But you are right, if you aren’t a serious collector but just collect for quantity or are no collector at all you won’t have these informations in your tags. I also don’t store any matrix /runout codes in my tags but I compare them to providers to get my exact match manually. What I store for instance is the release country, the label, the catalog number of the label, the barcode and the release date. This gets me pretty close and often is enough, for the rest I compare matrix codes. But I also admit this is really picky.

The problem here I see, as often mentioned now, is not Roon. I am sure you can get that. It is : Do your providers have the information about these releases…

At the moment I can see also problems with compilations by Mix-DJs like mentioned here by
@freelancr

I have also a lot of them where these problems, with big gaps in performers,… occur. There are really better sources out there but don’t know if considering such things is an option to you.

For me it would be not that bit problem to input that manually if it was possible, where I mean a certain release is groomed by you where the tracklist and all other things fit. So if then is just the release date, the label, a name extension (eg. 40th Anniversary edition) and country I would add that. It’s manual work again, but hey, it’s all in my tags already. Whereas name extension and album art are already possible.

So might it be possible to get a field release country (like the field Edit Version) in the album details, and a field catalog # maybe. These would be really great. Theses can really be plain text fields if you don’t want to use it for aquiring any rich data.

As I am developer myself I fully understand and respect this point.

This is of much more interest for me because my metadata is full, up to, or down to the barcode of nearly every release I have. I am really looking forward to what you will bring us there!!!

You achieved a lot already and hopefully this is going on. Better take the time you need before putting out something buggy. I would have only had concerns if you had told me come back and ask in 2 years :wink:

Thanks for getting back to me and letting me know of your thoughts on this.

@brian : One more thing. Are you interested, or better, does it make sense to post issues with metadata for certain albums on a more detailed basis here in the forum? I don’t mean issues that we discussed here like exact release match or so. I mean issues where the albums are identified but there are performers missing (eg like in the mentioned mix compilations), wrong track names, wrong album names, original release dates = later than release
If yes, should that be done on a one thread per issue basis or on a more collective thread?

What I store for instance is the release country, the label, the catalog number of the label, the barcode and the release date. This gets me pretty close and often is enough, for the rest I compare matrix codes. But I also admit this is really picky.

Can you go into some more detail on exactly how you store this information in your collection?

The problem here I see, as often mentioned now, is not Roon. I am sure you can get that. It is : Do your providers have the information about these releases…

Our biggest providers of release data today are Rovi (AllMusic) and MusicBrainz.

These two providers have very deep data models, but are not necessarily as broad as other providers (like Gracenote, which covers more releases, but provides much less detailed data about them).

Rovi has information about many, many releases, and certainly more than you can see in our UI.

One of the problems is: they don’t have track listings for all releases. Many of them just have a very small entry that acknowledges a release date, label, and catalog number/upc/etc. We don’t currently do a lot with these entries because so much of our backend is focused on using track listing similarity (both text and track lengths) to help with identification, grouping related releases, and so forth.

We have been discussing working with or incorporating data from Discogs, which I think does have a lot of good release data that we don’t currently have. From a technical standpoint, integrating something like this is straightforward assuming that we can get regular dumps of their database to run through our ingestion process.

I have also a lot of them where these problems, with big gaps in performers,… occur. There are really better sources out there but don’t know if considering such things is an option to you.

We just made some improvements (not yet released) to this. We are going to be looser about when we include/merge track credits data from tags with data from metadata sources.

So might it be possible to get a field release country (like the field Edit Version) in the album details, and a field catalog # maybe. These would be really great. Theses can really be plain text fields if you don’t want to use it for aquiring any rich data.

I can look into it. The chance of incorporating fields like this goes way up if they exist in any of our metadata providers (they provide more value to more users if we can pre-populate them).

1 Like

Sure I will provide you details about this, will write a somewhat detailed post a little later

Regarding AllMusic I said a lot in the last days, I also have an AMG implementation in my ripper, so I already see what they are providing at least in terms of track names and artists. I am not amused, they might be very good for certain music genres but they often have big problems with correct spelling and they very often lack in full track names, especially in mix-CDs (lacking the name of the mix, the performer etc.) Sometimes even freedb (which is also really painful in terms of spelling sometimes) has really better results than they have.

I realized the problems with no tracklistings when looking for some of my releases. Ok, so now I understand a little better why you have really little releases for some albums in your search list.

Didn’t know that you already work with a free provider like MusicBrainz. From my personal point of view and my experience discogs is really better in terms of the richness of releases they have and also in the correctness of the data. They also provide really helpful information in many of the release notes if the tracklisting or -naming is wrong on the cover. I mentioned I have only one release in my collection I cannot find at discogs, barcode and everything right. This is really great. I think a discogs integration can only be an enrichment to the data quality.

Ok so we should wait on the results after this is released.

I see one big problem with that if you take it from the metadata providers. If we have no really exact release matching as we discussed before then these fields propably won’t be right.
Or do you mean just pre-populating it on editable fields? Then forget my concerns.

Or do you mean just pre-populating it on editable fields? Then forget my concerns.

I mean pre-populating it on editable fields.

1 Like

Yes, if you read carefully you can read it out of your sentence. So this would be really nice

So, now as promised a bit more detail on my metadata in my collection :
Primarily I am using a highly customizable tagger that uses discogs API.

What I just normally have from my ripping process is
ALBUM, ARTIST, ALBUMARTIST, TITLE, DISCNUMBER, TRACKNUMBER, TRACKTOTAL, DISCTOTAL, GENRE, DATE

With this basic information and the original media of course I use the mentioned tagger looking for the correct release with album artist, album title, label, media type, catalog #, release date. If I find only one release I can tag, if there are more releases with this information (eg same catalog number, year etc.) I am looking deeper into release country, media type (eg special edition) and if needed barcode. That does it in most cases. If this is really still not enough I am using the matrix / runout codes of the media and check manually.

If I have the right release I tag the following (all through the tagger which uses the API of course) :
First as an explanation for Track and Release credits :
The release credits which are mentioned in discogs only for certain tracks (these are NOT track credits in discogs, we have separate track credits) eg.
Vocals - XYZ (tracks 1-9) are automatically appended as Track credits from the tagger

ARTIST
TITLE
TRACK HEADING track heading of discogs (eg “CD 1” or “Part I”) if the media is splitted into sub-parts
ALBUM
DATE
RELEASE YEAR
RELEASE MONTH
RELEASE DAY
ORIGINALYEAR release year of the master release
GENRE
STYLE
COMPOSER : (merged track credits and release credits, then filtered the roles ‘Composed By’,‘Score’,‘Written-By’,‘Written By’,‘Music By’,‘Programmed By’,‘Songwriter’ and applied the names that belong to the given roles as tags)
PERFORMER : (same as composer with ‘Performer’,‘Performed’)
LYRICIST : (same as composer with ‘Lyrics By’,‘Words By’)
VOCALS : (same as composer with ‘Vocals’)
REMIXED BY : (same as composer with ‘Remix’)
FEATURING : (same as composer with ‘Featuring’,‘Vocals [Featuring]’,‘Rap [Featuring]’)
DJMIXER : (same as composer with ‘DJ Mix’)
PRODUCER : (same as composer with ‘Producer’,‘Arranged By’,‘Recorded By’)
INVOLVED PEOPLE : all track credits merged with all release credits but the fields for PERFORMER, LYRICIST etc. filtered out because they are in the separate tracks already
COMMENT : Release notes
STYLE : Release styles
COMPILATION : compilation info indicator
LABEL : release label(s) (unique if there are dups)
CATALOGNUMBER : release labels catalog number(s) from discogs (unique if there are dups)
COUNTRY : release country
MEDIA : media format in the form : release formats quantity - release formats name - release formats description (eg Special/Limited Edition, Digipak ,…), so eg. “1 x CD, Album, Limited Edition, Digipak”
BARCODE : release barcode
ARTISTS ALIASES all available aliases of the artist
ARTISTS MEMBERS all members if artist is a group
ARTISTS IN GROUPS in which groups does the artist(s) play
ARTISTS ALL NAME VARIATIONS
ARTISTS URLS http-sites of the artist(s)
ARTIST REALNAME
ARTIST PROFILE profile description of the artist

So that’s my structure pretty much, there are a lot of technical fields like ACCURATERIPID, ENCODER, … but this is not really relevant I think…

If needed I put my hands on it manually eg for editing the label and cat.no information if there is more than 1 entry, this is often the case because there is label and distributor there. Sometimes editing COMMENT but don’t need to edit the other information very often. This is pretty accurate for my needs.

So it will look something like that (cutted out technical fields)

ALBUM=Anastacia
ALBUMARTIST=Anastacia
ARTIST=Anastacia
ARTIST PROFILE=Anastacia is an American singer-songwriter born on September 17, 1968. Anastacia has been highly successful in Europe, Asia, South Africa and South America, but has not had as much success in her native United States. Her debut album, “Not That Kind”, released in 2000, achieved multi-platinum sales in Australia, New Zealand and much of Europe. This resulting in her being critically acclaimed by both Elton John and Michael Jackson.
After having the 4th biggest European single and biggest Australian single of 2000 she was awarded “Best-selling International Artist” in 2000. Her next two albums continued her multi-platinum selling success throughout Europe making her one of the fastest and biggest selling artists of the new millennium.
In 2005, she was recognized for worldwide sales of over 20 million albums/records. Anastacia is world-renowned for her powerful soul voice, and her small stature, 5 feet 3 inches or 160 cm, this has led to her being nicknamed “The Little Lady with The Big Voice”. She is also known for her trademark glasses (often darkly tinted)…
ARTIST REALNAME=Anastacia Lyn Newkirk
ARTISTS ALL NAME VARIATIONS=A. Newkirk, A. Newkirk-Jive, Anastacia L. Newkirk, Anastacia Newkirk, Anastascia, Anastasia, Anastasia Newkirk, Newkirk
ARTISTS URLS=http://www.anastacia.com, http://www.facebook.com/Anastacia, http://instagram.com/cofapt#, http://anastaciaofficial.tumblr.com/, http://twitter.com/AnastaciaFanily, http://en.wikipedia.org/wiki/Anastacia, http://www.youtube.com/anastaciavevo
BARCODE=5099751347125
CATALOGNUMBER=EPC 513471 2
COMPOSER=Anastacia
COMPOSER=J. Rzeznik
COMPOSER=K. Dioguardi
COUNTRY=Europe
DATE=Mar 29 2004
DISC=1
DISCNUMBER=1
GENRE=Rock
INVOLVED PEOPLE=A&R: David Massey
INVOLVED PEOPLE=Art Direction, Design: Julian Peploe
INVOLVED PEOPLE=Coordinator [A&R]: Evan L
INVOLVED PEOPLE=Executive-Producer: Anastacia, Lisa Braudé
INVOLVED PEOPLE=Management [Artist Management]: Lisa Braudé
INVOLVED PEOPLE=Mastered By: Ted Jensen
INVOLVED PEOPLE=Other [Hair]: Robert Vetiga
INVOLVED PEOPLE=Other [Makeup]: Francesca Tolot, Sharon Gault
INVOLVED PEOPLE=Other [Styling]: Ligia Morris
INVOLVED PEOPLE=Photography By: Isabel Snyder
INVOLVED PEOPLE=Arranged By [Vocal]: Siedah Garrett
INVOLVED PEOPLE=Backing Vocals: Anastacia, Audrey Wheeler, Lisa Vaughn, Siedah Garrett
INVOLVED PEOPLE=Bass: Lance Morrison
INVOLVED PEOPLE=Drums: Matt Laug
INVOLVED PEOPLE=Engineer [Assistant Mix]: Keith Armstrong
INVOLVED PEOPLE=Engineer [Assistant]: Anthony Kilhoffer, JD Andrews
INVOLVED PEOPLE=Guitar: Tim Pierce
INVOLVED PEOPLE=Keyboards: Glen Ballard
INVOLVED PEOPLE=Mixed By: Chris Lord-Alge
INVOLVED PEOPLE=Producer [Lead Vocal]: Ric Wake
INVOLVED PEOPLE=Recorded By [Vocals]: Thomas R. Yezzi
INVOLVED PEOPLE=Vocals [Hypeman]: Mc Foureyez
LABEL=Epic
LYRICS= %%%% CUTOUT FROM ME NOW %%%%
MEDIA=1 x CD, Album, Copy Protected, Enhanced
ORIGINALYEAR=2004
PRODUCER=Glen Ballard
PRODUCER=Bill Malina
ORGANIZATION=Epic;Epic;Daylight;Daylight
RELEASE DAY=29
RELEASE MONTH=03
RELEASE YEAR=2004
STYLE=Pop Rock
TITLE=Seasons Change
DISCTOTAL=1
TRACKTOTAL=12
TRACKNUMBER=01
YEAR=2004
COMMENT=© 2004 Sony Music Entertainment Inc. / ? 2004 Sony Music Entertainment Inc.

May carry several stickers.

1 - Universal Music Corp/Poho Productions (ASCAP)/K’Stuff Publishing/Corner Of Clark And Kent Music/EMI Virgin Music, Inc. (ASCAP)
Recorded at Record Plant, O’Henry Studios, Record One, Larrabee East, CA
Mixed at Image Studios, CA

2 - Universal Music Corp/Poho Productions (ASCAP)/EMI Blackwood Music Inc. obo itself and Crypton Music (BMI)/Aerostation Corporation/Universal - MCA Music Publishing a div. of Universal Studios, Inc. (ASCAP)
Recorded at O’Hentry Studios, Record One and Record Plant, CA
Mixed at South Beach Studios, Miami Beach, FL

3, 4 - Universal Music Corp/Poho Productions (ASCAP)/Crypton Music/EMI Blackwood Music, Inc. (BMI)/Aerostation Corporation/Universal - MCA Music Publishing a div. of Universal Studios, Inc. (ASCAP)
Recorded at O’Henry Studios, Larrabee East, Record Plant, Record One, CA
Mixed at South Beach Studios, Miami Beach, FL

5 - Universal Music Corp/Poho Productions (ASCAP)/Sony/ATV Sounds LLC/Turtle Victory Music (SESAC)
Recorded at Record Plant, Record One, and Larrabee East
Mixed at Image Studios

Vocals recorded at Henson Studios
6 - Universal Music Corp/Poho Productions (ASCAP)/K’Stuff Publishing/EMI Music Publishing Ltd./Reverb Music Ltd admin, by Songs Of Universal, Inc/Danny Weissfeld Designee (BMI)
Additional production for W&R Group

Recorded at Cove City Sound Studios, NY, The Hit Factory Criteria, Miami, FL
Mixed at Cove City Sound Studios, NY
7 - Universal Music Corp/Poho Productions (ASCAP)/K’Stuff Publishing (BMI)/WB Music Corp/Dylan Jackson Music (ASCAP)
Recorded at Henson Recording Studios, Los Angeles, CA
Mixed at South Beach Studios, Miami Beach, FL

8 - Universal Music Corp/Poho Productions/Universal-MCA Music Pub a div. of Universal Studio, Inc/Aerostation Corporation (ASCAP)/K’Stuff Publishing (BMI)
Recorded at O’Henry Studios, Record Plant, and records One, CA
Mixed at Image Studios

9 - Universal Music Corp/Poho Productions (ASCAP)/Eligible Music Ltd. (PRS)
Recorded at Sony Studios, NYC and Neptune Valley Studios, LA
Engineered at Unstable Studios
Vocals Engineered at Sony Studios, NY

Mixed at Image Studios, CA
10 - Universal Music Corp/Poho Productions/K’Stuff Publishing (BMI)/WB Music Corp/Dylan Jackson Music (ASCAP)
Recorded at Henson Recording Studios, CA
Mixed at Image Recording, CA

11 - Universal Music Corp/Poho Productions (ASCAP)/K’Stuff Publishing/Bumyamaki Music/EMI April Music Inc. (ASCAP)
Recorded at Cello Studios, CA
Mixed at South Beach Studios

12 - Universal Music Corp/Poho Productions (ASCAP)/Eligible Music Ltd. (PRS)
Recorded at Sony Studios, NYC and Neptune Valley Studios, LA
Mixed at Images Studios, CA
Lead vocal production for Make It Rock Productions, Inc. for tracks 1 to 4, 6, 8, 9, 11
Mastered at Sterling Sound, NY

Wow!

Whilst Machine clearly has very comprehensive metadata which suit his (because it really has to be a boy! - no offence intended) particular needs, if Roon does adopt this metadata template, I sincerely hope it keeps most of that detail very well hidden from (my) view!

If my use of Roon involved ploughing through that level of (what I consider to be) detail surplus to requirements, it would actually compromise my experience to the point where I would throw in the towel and seek an application with a more normal level of metadata information.

Machine - that comment really isn’t intended to belittle your sincerity or requirements in any way. It is merely intended to inform the Roon team that not every user finds that level of information either helpful or desirable.

1 Like

@neilf : I really don’t think Roon is considering to adopt my complete metadata template :wink: So I don’t think you have to be afraid of that. Brian just wanted information on how I store my info…

@neilf, We do a lot of work to hide complexity, or put it in proper context. I don’t see that changing.

1 Like

Hey @Machine – no harm. Ronnie’s post and mine were placed at the same time - no double whammy intended. ‘Eclectic’ just refers to the width and scope of your collection and workflows. As I said before: you raise some valid and well thought-out points.

C’est le ton qui fait la musique. :wink:

1 Like

btw, much of this metadata is already in roon, but handled somehow in the “roon experience”. If it wasn’t there you would not be able to do what you can do in Roon already

and, @neilf : this is a release that has extremely detailed comment field, this is not necessarily always the case, that blows it up a little

Thanks for your reply!

Agreed on that :wink:

1 Like

@brian: I think this is a very good idea. Discogs provides a monthly data dump of their DB:

This data is made available under the CC0 No Rights Reserved license:

3 Likes