What I mean is…
If an artist entry gets populated with more than one artist of the same name, particularly if it has sat there for a length of time…
Instead of tracking down the original intended artist, and removing everything else - Do you think we should put all of the information (including the original intended artist) into new artist entries, and have the original link say something like “you may be looking for one of the following artists”.
I don’t want what Wikipedia does, where they have an ever-growing list with all “John Smith” articles on one page. I would want just the artists that were mixed into that one entry.
Right now, I am looking at a mess of a page, where the original artist has been hijacked by 300 edits of another artist - which includes linking to Wikidata and vice versa.
I think if we were to have a “correction page”, acknowledging that at one point in time an entry contained information on multiple artists, but we have taken steps to correct it and to please update your link with one of the following artists which shared the link.
I think that would do a lot of good.
And it would certainly help credibility. If, let’s say, I was on Wikipedia and clicked the MB link. Instead of finding ‘obscure artist John Smith’ when I wanted ‘ultra famous mega star John Smith’.
As I understand things the current approach would be to create a new artist for every identifiable John Smith artist, and make a disambiguation for each, as well an annotations , 1. on the originating multiple artists John Smith page warning users of the changes, and 2. if there are suspected links between John Smiths.
What you propose, if I understand you correctly, is a sort of mega-annotation that warns people that there have been significant changes around just who is included in John Smith (original).
I know a lot of time can be spent for little gain when the db is misread by this user.
We have had a few cases where there was say “Dave” and then a billion Daves in there, and at least an intermediate solution was to split the ones we knew, and turn the old one into something like “Dave (unknown Daves, please help identify)”. Ideally we shouldn’t need those eventually, but
I am also thinking that the original page would be removed from search results, nor would anyone be able to edit it. It would only be available for someone who has the link saved somewhere. Sort of the same way as when we merge two artists - we keep the original link active.
This way, if other databases, particularly the identifers (ISNI, VIAF, LoC, etc) link to an item that we messed up, someone isn’t being led astray.
Maybe all of the text is getting mumbled, so let’s try demonstrating an example -
MBID “A” is listed as John Smith (1).
After looking at MBID “A”, it is determined that there are 3 people named John Smith that are all showing at MBID “A”.
I think we should create John Smith (2), John Smith (3), and John Smith (4).
And MBID “A” would become John Smith (splash page), with a statement of -
The data at this entry has been separated into John Smith (2) at MBID “B”, John Smith (3) at MBID “C”, and John Smith (4) at MBID “D”. If you are seeing this message, please update your link to the correct artist.
Old Artist: John Smith page users are warned that they may have bad data.
They are offered link/s to good/better data.
The time-saving/corrupting option of just chucking all John Smiths in together is made more trouble for delinquent editors.
The existence of slightly differentiated Artist:John Smith (genre, release, track, role, relationship) makes it easier for, and more likely that, future editors, even lazy ones, will attempt to select the correct John Smith - because they won’t have a catch-all bucket Artist available even if they’ve used on in the past.
It will be easier for people checking edits to see whether data added to Artist: John Smith(robot drummer) matches that specific John Smith as all Artist: John Smiths will be checked now for accurate disambiguating relationships.
It depends on how many cases of this there are. It sounds nice idea for those who look at the MusicBrainz site with human eyes and brains to decode.
But the one area I would be a little concerned by are those who access MB through the Artist MBID’s they downloaded on a previous time. I’m a KODI user and Artist MBID’s are stored in the KODI database to speedup the lookups both here at MB and the lookups into many other databases online that key from the Artist MBID
I know nothing about how the API works, but if a page is shown to humans that says “John Smith (splash page)” the humans can then see they have to make a choice - how would that then handled by the API?
Will something special be added to the API to return an error of some form? Or a “You need to be more specific with that” error? Or a “Check your John Smith’s” error? Or just return the most popular John Smith for the original MBID?
I do also realise that this issue is more about those very minor artists. Like the guy who played the trumpet on that obscure track. That kinda implies that this should be a lower impact change even if the API just returned the most popular John Smith.
