[critiquebrainz] alastair merged pull request #307 (master…CB-369): CB-369: Write a review page shows incorrect information depending on if the item is a work or artist https://github.com/metabrainz/critiquebrainz/pu...
_lucifer: this uses the production database, but please have a click around and see if you find any bugs (stars look OK here)
especially focus on the things that you changed
_lucifer
sure, will do!
ruaok
alastairp: anything for us to discuss today?
yvanzo
ruaok: Would it be alright to create a node on Hetzner Cloud to further test indexer load issues?
ruaok
sure.
do you know what size you need?
yvanzo
Not exactly yet.
alastairp
ruaok: don't think so. I have 1 more half-day feature that I need to finish on the similarity API before releasing it, that'll be done by the time we meet on Weds
ruaok: also, I was thinking, should we just remove score from recommendation card, and just display them in a random order? The score is rather unattractive
shivam-kapila: but then they will be stored in a different db?
shivam-kapila
No
But why do you want to store in different db
pristine___
Where will the feedback be stored?
shivam-kapila
reosarevok: there isnt a page rn. My bad
pristine___: recording_feedback table
pristine___
shivam-kapila: To use it to refine recs?
shivam-kapila
Why not use entire feedback
ruaok
pristine___: I'm ok with that. how about we add the love/hate buttons, but have a tooltip/alt/whatever that shows the track ranking if you want to know what it is?
pristine___
ruaok: we can later when the ranking actually has a meaning.
ruaok
it doesn't now?
shivam-kapila
[-1,3]
Is vague imo
pristine___
Not really. The range belongs to (-1, 3). This is what I have got from all the runs of the rec scripts. This range does not really make sense to me, also, like today you have only one score
For all your similar artist
ruaok
wut?
are those observed values?
pristine___
We have scaled the ratings.
shivam-kapila
iliekcomputers: the work is done. I will push 2mrw
Sorry for delay
ruaok
yes, I know we scaled the ratings. is the [-1,3] after the scaling or before?
pristine___
Before
ruaok
rather than get rid of the score we need to improve the score to make it worthwhile.
I guess they are not fixed to a particular scale -- this means that the scale is somehow natural to our setup.
pristine___
Right. I am not saying get rid of them, I am just saying till we make them worthwhile we may consider not showing them to the users
ruaok
if we change something, then the scale changes. which blows.
lets make them worthwhile.
shivam-kapila
Imo it should define the likeliness of track to match my taste
ruaok
and then way to do that is to scale each user independently.
pristine___
Yes, the scale really depends on the input which is right now the play count (listen count)
ruaok
find the max and the min for each user and then [min, max] -> [0,1]
pristine___
_lucifer: is working to normalize the input
> find the max and the min for each user and then [min, max] -> [0,1]
ruaok
what is _lucifer *not* working on?? :)
what does that get us?
pristine___
This will only scale the ratings. I am trying to get more meaningful output from the recommender itself. 0.99999999988887 and 0.99999999999886 doesn't really make sense to me.
> what does that get us?
^
ruaok
ok, which part exactly are we scaling? there are many inputs
I’ve received a single mailed-in review from zas, so… zas: Go!
zas says…
"""
This week I mainly worked on Picard, PRs reviews.
Plus usual stuff, admin tasks, upgrades, supervision, user support.
I'll prolly upgrade Grafana to 7.2.x this week, it has some breaking changes I need to review first.
303 crawler issues are more or less solved (I blacklisted most of their IPs), it was using quite a lot bandwidth on MB website last weeks.
Among bad news, I was expecting to be at OfficeBrainz for the summit, but I had to change my plans :( trip was cancelled.
"""
Others up tonight: yvanzo, reosarevok, pristine___, bitmap, ruaok, alastairp, Freso, Mr_Monkey, CatQuest, shivam-kapila, _lucifer, jmp_music – anyone else who want to give review, let me know! :)
yvanzo: Go!
yvanzo
Hi!
Last week, a new version of MusicBrainz Server has been released.
We announced a breaking change to fix inconsistencies in JSON response from web service.
(small inconsistencies between lookup/browse and search queries)
More are being fixed by _lucifer :)
reosarevok
!m _lucifer
BrainzBot
You're doing good work, _lucifer!
Mr_Monkey
+1
yvanzo
I finally unlisted the thousands of edits from MBS-10921.