might also give an example of an actual working URL
2020-09-21 26509, 2020
yvanzo
bitmap: it should be possible to use link.entity.gid with recording_gid_redirect instead of just link.entity.id, but that is probably far more complex.
2020-09-21 26537, 2020
pristine___
Yeah, that was my intention, people may find <user-name> thing weird
Found a small easy-to-fix bug in the YoutubePlayer ^
2020-09-21 26540, 2020
Mr_Monkey
Nothing about that "missing parameter type" error message yet, though.
2020-09-21 26502, 2020
Mr_Monkey
Nicely done pristine___, shivam-kapila, ruaok !
2020-09-21 26502, 2020
Mr_Monkey
Really cool and exciting to see recommendations pages in LB ! :heavy_heart_exclamation_mark_ornament:
2020-09-21 26516, 2020
ruaok
:D
2020-09-21 26523, 2020
Mr_Monkey
Feels like a really important step.
2020-09-21 26550, 2020
ruaok
let's see what the community reaction will be.
2020-09-21 26509, 2020
pristine___
Mr_Monkey: thank you :)
2020-09-21 26558, 2020
abhinavohri joined the channel
2020-09-21 26539, 2020
Mr_Monkey
Same conclusions for me: 1) the playlists are on point, and they are indeed tracks I might very well like and 2) they do need more diversity
2020-09-21 26539, 2020
Mr_Monkey
Really good first version though!
2020-09-21 26535, 2020
pristine___
Thank you Mr_Monkey for the feedback.
2020-09-21 26544, 2020
pristine___
Did you get the chance to read the community blog?
2020-09-21 26557, 2020
pristine___
You will understand better the reason for point 2.
2020-09-21 26509, 2020
pristine___
But we are working in the diversity :)
2020-09-21 26512, 2020
pristine___
On*
2020-09-21 26539, 2020
Mr_Monkey
Reading now :)
2020-09-21 26546, 2020
ishaanshah
pristine___: Hi, the recommendations are really cool, nice job :D
2020-09-21 26557, 2020
shivam-kapila
> why are there affordances that tell me any of this?
2020-09-21 26503, 2020
ishaanshah
I was reading the your community post right now
2020-09-21 26505, 2020
shivam-kapila
ruaok: didnt get you
2020-09-21 26534, 2020
ishaanshah
> Even if you have listened to an artist many times, it is highly probable that the tracks associated with that artist will not show up in your playlist given the artist is not popular in ListenBrainz.
I wonder if there will be a significant self-influencing effect, ie if I listen to my recommendations a lot, will that generate more similar recommendations, possibly less diverse ones over time?
2020-09-21 26547, 2020
ishaanshah
I am asking this because I recently started listening to some country music from a very less popular artist but I would like to get more artists similar to her....
ruaok: ah gotcha. we had a hover title for the cards saying "Double click to play. " but iliekcomputers suggested its too much
2020-09-21 26538, 2020
pristine___
ishaanshah: we can take recordings from some big data set which is available for public use
2020-09-21 26540, 2020
pristine___
Maybe
2020-09-21 26540, 2020
ruaok
and nothing at all is too little.
2020-09-21 26505, 2020
iliekcomputers
i did? i don't remember
2020-09-21 26523, 2020
ishaanshah
pristine___: something like mlhd?
2020-09-21 26556, 2020
pristine___
ishaanshah: that's not how CF works, we take recordings from data submitted to listenbrainz, that's why they too much depend on what the LB users listen to
2020-09-21 26509, 2020
pristine___
That's the biggest drawback
2020-09-21 26513, 2020
pristine___
ishaanshah: yes
2020-09-21 26520, 2020
shivam-kapila
what I was thinking is that we need a release banner which highlights the stuff we just released. A single component that shows up after major (or maybe even minor) features are added
2020-09-21 26541, 2020
shivam-kapila
ruaok: iliekcomputers ^^
2020-09-21 26512, 2020
ishaanshah
Hmm, so if I am the only one to listen to a particular artist on LB, then theres no way that I'll get similar artist to that artist?
2020-09-21 26547, 2020
pristine___
ishaanshah: I mostly liisten to Indian artists , but my recs are not diverse because there aren't much listens in LB that belong to Indian artists, it really depends on users and the amount of data we have rn.
2020-09-21 26548, 2020
iliekcomputers
The tweet does the same thing
2020-09-21 26555, 2020
iliekcomputers
It shows up on the home page
2020-09-21 26503, 2020
pristine___
> Hmm, so if I am the only one to listen to a particular artist on LB, then theres no way that I'll get similar artist to that artist?
2020-09-21 26505, 2020
abhinavohri
i want to work on LB-583. Can someone provide me some details about the test add where to add it.
2020-09-21 26506, 2020
pristine___
You will get a similar artist, but you might not get recordings of that similar artist if no user of LB has ever listened to that artist.
ishaanshah: the best option is to use a bigger pool of data like mhld or something. People seldom want to search manually for tracks even if they have a recommend artist ig
2020-09-21 26532, 2020
pristine___
0.598 shivam-kapila
2020-09-21 26532, 2020
_lucifer
one thing that could be done is that we source the entire recording dataset mb has, maybe give each recording a value of 1
2020-09-21 26534, 2020
pristine___
Yes?
2020-09-21 26508, 2020
shivam-kapila
I have a suggestion that the score should be normalised per user rather than based on whole data set
2020-09-21 26533, 2020
_lucifer
yes shivam-kapila, i agree
2020-09-21 26541, 2020
shivam-kapila
having a rec with score 0.5 gives me a feel that I am getting half quality recs
2020-09-21 26512, 2020
pristine___
Using MB dataset, _lucifer doable, sounds good.
2020-09-21 26523, 2020
pristine___
Right shivam-kapila
2020-09-21 26530, 2020
pristine___
_lucifer: do you want to work on this?
2020-09-21 26535, 2020
shivam-kapila
Plus 1 more thing
2020-09-21 26539, 2020
pristine___
I remember you once said that you want to
2020-09-21 26541, 2020
_lucifer
sure, pristine___
2020-09-21 26555, 2020
shivam-kapila
how is the input scaled
2020-09-21 26504, 2020
pristine___
It's not, as of now.
2020-09-21 26504, 2020
_lucifer
it isn't iirc
2020-09-21 26517, 2020
pristine___
It's just the playcount.
2020-09-21 26533, 2020
_lucifer
scaling is a subset of normalizing
2020-09-21 26555, 2020
pristine___
_lucifer: I think you opened a ticket for this? Normalizing the input?
2020-09-21 26555, 2020
shivam-kapila
I have a feel if we scale it on basis of how the listens hold a portion of my listening habits we might imrpove results
2020-09-21 26505, 2020
shivam-kapila
_lucifer: I am poor at maths
2020-09-21 26525, 2020
_lucifer likes math ::sunglasses:
2020-09-21 26557, 2020
_lucifer
pristine___: let me see if i can find it otherwise i'll open one now
