I jumped back into working with it and installed the code in a different place and it just magically worked this morning (because you added a module capability)
2020-05-15 13624, 2020
ruaok
that was nice synchronicity!
2020-05-15 13625, 2020
ruaok
thanks!
2020-05-15 13600, 2020
rdswift
Glad to hear it's working for you. I plan to make a small upgrade to the module sometime soon to allow adding lines to the header section if required, and to also add to the message body.
MBS-10828: musicbrainz-docker createdb.sh fails.. out of space
2020-05-15 13608, 2020
Mr_Monkey
Implementing multiple music sources is a huge pain… but it'll be worth it !
2020-05-15 13636, 2020
ruaok
I hope you're making it nice and modular so we can add more sources
2020-05-15 13640, 2020
jmp_music has quit
2020-05-15 13645, 2020
Mr_Monkey
That's the idea.
2020-05-15 13653, 2020
ruaok
<3
2020-05-15 13603, 2020
Mr_Monkey
I'm gutting the poor spotify player out
2020-05-15 13613, 2020
Mr_Monkey
There's lots of back and forth between the components, so lots of handler to pass around, but It'll be modular all right
2020-05-15 13655, 2020
Mr_Monkey
(The main player component has the playback controls, while each music source can update progress, pause/play state, etc.)
2020-05-15 13602, 2020
Mr_Monkey
In the future, it would be nice to implement feedback about incorrect song matching —maybe with an option to provide a better match— to post some info to MessyBrainz
ishaanshah[m]: a compromise here would be to do the filter in python and then get the actual stats in sql
2020-05-15 13643, 2020
alastairp
I took the listenbrainz artist stats, and some code that one of my coworkers at UPF had written, using matrix factorisation colaborative filtering
2020-05-15 13615, 2020
iliekcomputers
i'm very hesitant to accept ORM code when all the rest of our code uses SQL
2020-05-15 13625, 2020
ishaanshah[m]
Hmm, that sounds reasonable
2020-05-15 13634, 2020
ishaanshah[m]
I agree that SQL is more readable
2020-05-15 13644, 2020
alastairp
it generated this model. given the "train" data, it made these recommendations. The "test" data was held out of the list of data that was given to the model, but we know that the user listened to it. The True/False in the Recommended line is if the recommendation exists in the test set or no
2020-05-15 13602, 2020
alastairp
this data looks really good. I think I can look up recs for a given username too, with a bit of work
2020-05-15 13605, 2020
iliekcomputers
because eventually one or two years down the line, we will have to migrate to one method because it's too complicated.
2020-05-15 13617, 2020
iliekcomputers
which is something I would like to avoid.
2020-05-15 13621, 2020
ruaok
alastairp: one wrong? not bad.
2020-05-15 13627, 2020
alastairp
in addition to this, it spits out artist similarities. here are similar artists to the beatles, using a few different metrics:
2020-05-15 13633, 2020
alastairp
SIMILARS ['The Beatles', 'The Rolling Stones', 'Electric Light Orchestra', 'Simon & Garfunkel', 'Arcade Fire', 'Bob Marley & The Wailers', 'John Lennon', 'The Who', 'Paul McCartney', 'Oasis']
2020-05-15 13633, 2020
alastairp
SIMILARS ['The Beatles', 'John Lennon', 'Simon & Garfunkel', 'Paul McCartney', 'George Harrison', 'The Rolling Stones', 'The Hollies', 'Electric Light Orchestra', 'The Mamas & the Papas', 'Arcade Fire']
2020-05-15 13646, 2020
alastairp
ruaok: right, and even then "wrong" isn't so much the correct word to use here
2020-05-15 13650, 2020
ishaanshah[m]
Yes, I agree
2020-05-15 13600, 2020
alastairp
it's just saying "the algorithm suggests these artists, but you haven't listened to it before"
2020-05-15 13606, 2020
alastairp
so that's a _good_ recommendation!
2020-05-15 13607, 2020
ruaok
let me generate my similar artists to beatles, hang on.
2020-05-15 13616, 2020
ishaanshah[m]
I will filter using ORM and then revert to the old query
