Non-English Metadata Range in 2023 (Spotify, Soundcloud, Bandcamp)

[davidmacfadyen]

Hi All ~
I have used SongKong with much success and for long periods in the past – and have now come back to renew my membership for another big tagging task. So my question falls into the “please-bring-me-up-to-speed” category. Almost everything I have is from Eastern Europe and therefore outside the Latin alphabet, BTW.

Are the following assumptions correct in 2023?
1 - SongKong pulls exclusively from MusicBrainz and Discogs for its metadata? I ask because I’ve been experimenting with OneTagger, which is more designed for dance music, but when I give it a task relating to media I know is in Discogs, the program just flies through the scan in less than a second and claims nothing is there.
2 - SongKong does not offer the stylistic coefficients/scores from Spotify (danceability, valence, acousticness, etc etc)? Having any generic or formal categories expressed as numerals can be extremely useful, esp when it comes to visualizing traits of v large data sets.

Thanks!
David

PS: Having SongKong trawl Bandcamp and Soundcloud, too, would be v useful for me, but maybe the counterargument is that such tags are always self-defined / self-applied and therefore inconsistent and/or often absent. If those two sources are not part of SK’s plans, maybe you could suggest a second tagger to search those platforms - separately/afterwards?

[paultaylor]

SongKong uses MusicBrainz, Discogs and AcoustId. AcoustId database doesn’t just hold AcoustIds and links to MusicBrainz recordings it can also hold basic metadata about the songs and we can use this even if the song is not linked to a MusicBrainz recording. Acoustid is actually much larger than Acoustid and MusicBrainz, it has 66M unique AcoustIds compared to 17M MusicBrainz Recordings so this is very useful but Acoustid only stores the basic metadata (artist, albumartist, title, album, trackno, year)

We did look at adding beatport but they no longer make their api available. There is a general problem with using any api that it may be change or removed or impose license restrictions to prevent usage. With MusicBrainz, Discogs and Acoustid the data can be downloaded and then implemented in a database, this is what we do with the Albunack database so we then are in control and can provide optimized solution for SongKong and Jaikoz

There is maybe a way to download the data from bandcamp/soundcloud and add it to Albunack and that would be our preferred solution.

Also worth noting there are various scripts available to make it easy to import a release from sites such as Bandcamp into MusicBrainz, and then SongKong will pick it up when update the Albunack database.
If everyone just imported a few albums using a script from https://wiki.musicbrainz.org/Guides/Userscripts that would make a huge difference.

MusicBrainz has a project called AcousticBrainz and if a song has a link to AcousticBrainz we can get these metrics, SongKong already uses this and adds the following numeric fields (range 0 - 100)

• Mood Aggressive
• Mood Relaxed
• Mood Sad
• Mood Happy
• Mood Party
• Mood Danceability
• Mood Acoustic
• Mood Electronic
• Mood Instrumental
• Tonality

I’m not sure how accurate these measures are, but something to look at.

[davidmacfadyen]

Thanks: that’s v kind of you. Given the close relationship between Jaikoz and SongKong – and my desire to avoid doing two scans of a massive collection (close to 2M compositions, most of which are not in English) – it would obviously be ideal to get the best metadata in one fell swoop. Jaikoz is the more nuanced tool, it seems, so does that also pull in the numeric data of AcousticBrainz?

[paultaylor]

It does (both need Pro version) but the SongKong matching algorithm is the better one currently, and actually Jaikoz uses more memory the more songs loaded so would not be able to deal with 2M songs, whereas SongKong can. Generally I would always advise using SongKong in first instance, and then only using Jaikoz if required to resolve particular matches.