A way to set Jaikoz to pick the most frequent Acoustid source?

[garymac22]

So let’s use this page as an example;

http://acoustid.org/track/cd5a0b30-36aa-4af9-a31a-cbe67d46a177

The Acoustid Id has (had) many different songs attached to it.

When I ran autocorrect, it chose to tag as “Love Hurts” by Nazareth, which had one source indicating this was correct.

The real song is “Is This Love” by Bob Marley, which Acoustid has over 1500 sources indicating that this was the correct song.

Is there a way to set Jaikoz to pick the highest-numbered source to generate the tags that are most likely correct, (based on the number of sources that suggest it’s correct)?

[paultaylor]

Not yet, but there will be, does existing metadata favour Love Hurts, if so try removing metadata fields and retrying.

Such an option has already been added to Songkong called Ignore existing metadata when matching individual songs.

[garymac22]

The song was actually tagged as “All Out Of Love” by Air Supply. When I noticed it was wrong, (by playing it), I erased the title metadata and then ran autocorrect. Its second choice was “Love Hurts”.
After that, I headed to the Acoustid Id page and disabled all the wrong info so that I could finally have Jaikoz identify the proper track.

I’m glad to hear a feature to ignore current metadata is coming.

I’m working with a lot of DJ-issued CDs, so they are not as widely distributed… thus not going to fall into the “usual” spots.

Obviously no tagger is going to match 100% perfectly and obviously Acoustid Id/MusicBrainz has a lot of bad info fed into it. Because of this, it would seem most-effective if the tagging process could be set up to go with the results that are most likely correct based on the quantity of verifying sources.

Is it correct to say that in the current version of Jaikoz that if a song has zero metadata, that the tagging options with the highest source count would be chosen?

Do you think it would be beneficial to erase all metadata and have my entire library looked at from scratch?

I’m not concerned about maintaining album grouping, because like I said, I’m not working from traditional albums anyway.

My ideal result is have all songs tagged correctly and attached to the album of their original release.

Suggestions?

Thanks in advance.

[paultaylor]

Well existing metadata can help find the right recording, its only a hindrance when the existing metadata is incorrect, so I wouldnt really recommend removing all metadata.

I don’t think its guranteed to take the one with most sources but you could try, sorry I’m deep into working on new release which under the covers has changed a fair bit form the current version.

Seeing as you are track based anyway the best thing might be to use [b}Manual Correct From Musicbrainz[/b], this shows you the matches it has found and you can select the one you want to choose, by right-clicking and selecting Match Songs to Selected Release you can match all tracks that have been linked to a track on the same release to that release , so you dont have to check one song at a time

[garymac22]

I would suggest that a principle purpose of using Jaikoz is to have it correct files with incorrect existing metadata, not to go along with incorrect data and support it.

At the heart of the issue is, if I have an MP3 that is Bob Marley’s “Is This Love”, and somewhere along the way, a few people, (or even me), have mis-labelled it as “Love Hurts” by Nazareth, then I want Jaikoz to step in and put an end to the madness. I want Jaikoz to identify that it is labelled wrong and fix it.

I can’t think of a logical scenario where I would want a piece of software to identify a song for which thousands of sources suggest it’s a certain song, yet support it being tagged as another song just because a few sources have got it wrong.

If I incorrectly had the song tagged as Love Hurts, why would I want it to stay labelled as Love Hurts if literally thousands of sources suggest it’s a Bob Marley song?

In this example, (to help me get a handle on what goes into it), how does this given Acoustid Id come to have 15 different songs associated with it? How does a song get attached to fingerprints at Acoustid Id?

Thanks again for your help.

[paultaylor]

[quote=garymac22]
In this example, (to help me get a handle on what goes into it), how does this given Acoustid Id come to have 15 different songs associated with it? How does a song get attached to fingerprints at Acoustid Id?
.[/quote]

Well that is the problem, acoustid analyses attributes of a song to create an audio fingerprint and then matches this fingerprint to existing fingerprint and retrieves the acoustid or allocates a new acoustid.

The problem is that two songs that sound completely different can end up with the similar fingerprint and identical acoustid neither may be wrong see https://acoustid.org/stats . So that is one way where the existing metadata helps as it is fairly unlikely that the existing metadata is wrong AND it matches a wrong recording linked to its acoustid, unless the metadata was incorrectly added due to previous incorrect acoustid match.

Acoustid/MB Recording Ids can be submitted to Acoustid by tools such as Jaikoz matching songs by metadata, getting the acoustids and then submitting the pairing (Action:Remote Correct:Submit MusicBrainz/Acoust Id Pair)

But of course there can also be proper errors if incorrect pairings are submitted. And also there is many-many mapping between MusicBrainz Recording Ids and AcoustIds where in a perfect world there would be a one-one mapping.

I would say the principle aim of Jaikoz is to get your library in better shape, which means adding more metadata and matching to MusicBrainz/Discogs for songs with decent metadata as much as identifying incorrectly tagged songs. Im my experience songs are much more likely to just be missing any useful information rather than having completely wrong information.

But if you like why not try removing all metadata (Edit:Delete All Metadata) and see what results you get. What I can say is that Jaikoz will definitently make use of Acoustid NoOfSources more in the new version.