I mean, this was last changed almost 10 years ago, I think it was re: the original implementation, and we probably just never touched it since it was already implemented, heh
A note about the similar artist api, once you have the list of artists, there is not an easy way to retrieve a recording by them (maybe with a rating bias or by total listens) to create a playlist.
2023-05-26 14625, 2023
jasje
isabelxxx: left comment on youur listen service ticket. Check it out!
2023-05-26 14632, 2023
jasje
your*
2023-05-26 14643, 2023
isabelxxx
I do it by myself, using the user library though (since I have ratings or playcounts there), but whenever such artist is not found, it becomes too convoluted to find a track by that artist (I could simply take the first result on youtube though).
Eminem is the most similar artist according to LB model xd.
2023-05-26 14620, 2023
isabelxxx
I would say it needs a tweak
2023-05-26 14636, 2023
zas
bitmap, yvanzo: floyd, pink & aretha still need to be rebooted (to fully apply docker upgrades)
2023-05-26 14630, 2023
yvanzo
bitmap: I guess we need to switch PG master/slave ^ but we don't have a doc for it
2023-05-26 14647, 2023
mayhem
whoooo hooo, I love seeing PRs from aerozol !
2023-05-26 14623, 2023
yvanzo
bitmap: for aretha ^ is there any service (barman, json dump, ftp, pg, redis, rsync) to pay attention to before rebooting?
2023-05-26 14657, 2023
monkey
lucifer: Are you sure everything is fine for the form? Let me know otherwise and I can take a look.
2023-05-26 14640, 2023
monkey
isabelxxx: It's really nice to see the similarity data integrated in another project, nicely done :)
2023-05-26 14640, 2023
monkey
Indeed, we have some more tweaking to do. You'll notice we don't really use this data directly in ListenBrainz yet. Part of the reason is we are working on the UX for it in LB, and the other part is that it's still under construction, even though we've definitely already gotten somewhere.
isabelxxx: yes, the data needs tweaking, for sure. it still early times. we're going to improve the data with two more steps, but it will take some doing.
2023-05-26 14623, 2023
mayhem
for now, assume that we know about the problems and will work to fix them in the near future.
2023-05-26 14639, 2023
mayhem
as far as artist -> recording, you're looking for this:
we'll be building several popularity datasets, this is the first one.
2023-05-26 14602, 2023
mayhem
but, please be aware that anything served from datasets.lb.org is *experimental* and not guaranteed to be stable and fully available. once the data moves to labs.api.lb.org we guarantee stability and uptime.
2023-05-26 14622, 2023
mayhem
find for dev work, but don't push those APIs into production.
2023-05-26 14628, 2023
mayhem
*use of
2023-05-26 14642, 2023
mayhem
akshaaatt: thanks for the ML link -- will look.
2023-05-26 14604, 2023
lucifer
monkey: yes, all good for now.
2023-05-26 14611, 2023
isabelxxx
mayhem: Somehow I missed that one, my fault. I will check it, since it allows to post an array, it would work great with the simil artist output. Thanks!
2023-05-26 14615, 2023
monkey
👍
2023-05-26 14632, 2023
isabelxxx
And will take note about your warnings :) It's great anyway to be able to play with all this as is.
2023-05-26 14637, 2023
mayhem
isabelxxx: most welcome -- that is exactly what I built it for. :)
2023-05-26 14652, 2023
lucifer
for similar artists, there is another algorithm available: `session_based_days_9000_session_300_contribution_5_threshold_15_limit_50_skip_30`
2023-05-26 14653, 2023
mayhem
I'm tickled pink that someone is appreciating the data.
2023-05-26 14610, 2023
mayhem goes to hug the dataset-hoster that makes all this possible
2023-05-26 14641, 2023
isabelxxx
The second algorithm did not recommend Eminem, but now The Beatles appear hahaha It is what it is.
