List of obscure recordings that have many Acoustids linked to them, most/all of the links were invalid. They seem to fall into two groups, either duration is not set, or track title is a funny one likely to match many things or a combination of both
I think if we could add the duration to those recordings in the list missing them that may prevent this problem reoccurring, but I struggled to find the duration of the ones I checked, perhaps someone else could have a look ?
Argh! Looked at the first one, it is more āmatch by nameā issues. You can see how everyone has named them differently causing the matching chaos. Many are the episodes all shaken up. Really need someone who owns the set to hit them with a reset and then a clean upload with the fingerprints carefully checked.
Can you spot the one with the most Fingerprints? Then torch anything that only has a couple of Fingerprints uploaded? It is also very possible to have multiple AcoustIDs on that one though as these will be different people doing their own tapes from the Radioā¦
As to [unknown] by [unknown] - I donāt even need to click on them to know there will be thousands in there. These are the old CD rippers who didnāt have track names.
Iām staying well away from theseā¦ We canāt fix all the madness. Iāll just focus on music that is relevant to me and some related tangents which I have chance to cross checkā¦
Iāve been keeping track of these, trying to do my part cleaning up stuff too (I just opened up the Linkin Park hole and went āoh nooo!ā).
Two things;
Is there any way that Picard could prevent AcoustIDās being uploaded to really incorrect tags? Maybe work it on the reputation of that user in MB? (so many correct edits before they can submit these)
I wish that we could somehow clear these dead links on MB after theyre removed, after so many months or so - it looks sooo messy.
My simple thought is not allowing more that 300 AcoustIDs uploaded in one go. A bulk upload is clearly not being checked. Reputation would be good but adds a different layer of complexity and tracking. Something simple like a number of submissions stops that accidental press of the button by people who donāt know what it does.
At the same time we need to encourage those who do have good data to add more AcoustIDs so the good data stands out from the bad. A difficult balence.
The only reason I come up with that odd number is because at this moment I am working a release with 223 tracksā¦
But totally, if someone has uploaded 1,000 AcoustIDās in one session and theyāre all bad, then it would be nice to be able to get rid of them (the IDās) without labouring over each one.
I donāt think āreputationā is necessary, more just tracking when someone has uploaded, as you say, 300+ AcoustIDs in one session, and theyāve all been attached to āTrack 01ā etc! imo there is a threshold where you can say, ok this session (or user, in extreme cases) is not worth the few possibly valid IDs.
Thank you to you guys who are tidying up regardless though
Itās too much for me unfortunately, too easy to for someone to add 1mil broken ones with a click.
With my SongKong tagger the Acoustids/Mbid pairings are submitted automatically, but only when a complete album has been matched to a MusicBrainz album, its much harder to match to the wrong album than it is to incorrectly match one individual track, also I do submit the metadata whereas I dont think Picard does.
I think at both ends (acoustid and Picard) it would be sensible to ban pairs that are more than x percentage differ in length, there are over 5000 tracks in Acoustid match to a mb recording that is either over twice the length of the fingerprint or half the length, these are never going to be a good match.