Genre Cloud Much Smaller After Re-install

[Gray Mouser]

I had to do a complete re-install of Subsonic, Postgresql and MusicCabinet. Everything was a fresh install, I did not use any backups of the database or any settings. I have everything back to the way it was before, but my Genre Cloud is much smaller now after scanning my library. Many artists that used to be in say, "female vocalist" (now missing as a genre) do not show in any genre although they did before my re-install.

Is this just due to limitations imposed by last.fm? Will my Genre Cloud rebuild itself over time, or is there anything that I can do to try to get all of the last.fm genres back?

Thanks!

[hakko]

Maybe this applies? How big is your library? viewtopic.php?f=11&t=10457

[Gray Mouser]

I have just 77 artists in the library, so that doesn't seem to apply.

[hakko]

Oh, that's another story then. The reason for this is that last.fm tags aren't perfect in any sense - a lot of misspellings and worthless tags like "seen.live", "awesome" etc. But since there's a lot of value in them (if you manage to filter them), I've tried various techniques to identify "valuable" tags and only use those.

Now the current algorithm sets a threshold saying that only tags with at least X artists should be considered for going into the tag cloud. This makes sense for large libraries (with thousands of artists) but maybe it's too hard for a library of your size. I'll lower it a bit in the next release.

To put female vocalists into your tag cloud right now, you'd have to alter the database. Use the pgAdmin tool that came with your PostgreSQL install, and run this SQL command in your musiccabinet database:

Code: Select all

insert into library.toptag (tag_id) select id from music.tag where tag_name = 'female vocalists';

[Gray Mouser]

Ah thanks, very helpful... I am pretty database-savvy (MS SQL Server Anyway) but here is what I did:

Code: Select all

insert into library.toptag (tag_id) select id from music.tag where id not in (select tag_id from library.toptag);

That will put the entire universe of last.fm tags back into the genre cloud.

I will just weed out all the tags I think are either not useful or stupid. I must have created my previous library database with a set of toptag values that included some of the ones that are now not in your list.

[Gray Mouser]

Just a thought -- perhaps you could alter the algorithm to accept the top xx% of the tags applied to any given song in a user's library -- that would have allowed certain artists in my library that have few tags, but not really popular tags, to get assigned to the library. Every re-scan would apply the most popular xx% tags as needed... Just a thought. But again, thanks for the help, it worked like a charm. Now I just have to weed out all the stupid tags (after I did what I did, rather than just put that one tag in as per your code)

Now, if you implement the "specific folder to share per user" feature request and I will definitely contribute... I will likely contribute anyway though, as I really, really like the MusicCabinet Mod, even though I have been using it for only a few weeks now...

[hakko]

The current algorithm takes the top 250 most popular tags (popularity counted as number of artists tagged with it by a substantial enough number of users), with the limitation that at least five artists should be tagged with it. Feel free to experiment with the query and post improvements here! That's the best way to contribute.

Code: Select all

select * from
(select t.tag_name, tc.tag_name, occ.count, case when tt.tag_id is null then false else true end from music.tag t
inner join (select tag_id, count(tag_id) from music.artisttoptag att where tag_count > 25 group by tag_id) occ on occ.tag_id = t.id
left outer join music.tag tc on t.corrected_id = tc.id
left outer join (select tag_id from library.toptag) tt on t.id = tt.tag_id
where occ.count >= 5
order by occ.count desc limit 250) top order by 1;