#metabrainz

Mōrena! A bit all over the place here still

fish

Finished off all the tutorial videos for the YouTube channel, a big pile of them. Paused the Picard ones for now, just not knowledgeable enough

blub

:D

Sweet

The usual stuff otherwise. Digging into some wiki/guideline pages - currently disambiguations and broadcast, gearing up for a well overdue soundtrack styleguide draft

Go riksucks!

Hmm they’re not online, who’s left

If riksucks and yellowhatpro are still not around, only one left is vibhoo_24

If they are here :)

gooooooo vibhoo_24!

they are not

He is not around!!

Prolly away too

Ok

So

That completes the reviews!

mayhem, aerozol: wanna talk about the playlist stuff for a moment, or do we move on?

I was just wondering what’s being launched :)

we can chat after the meeting.

(Oh also, I wrote meeting notes for the first time, hopefully will be quicker this time)

Ok!

So, next one up is, er, me

:)

I was looking at the comments on what to do with LB dupe(ish) users

And we don't really know yet

Basically, we have a bunch of users who clearly have several accounts and listen to the same stuff (pretty much) with both

spambrainz?

I still don’t think there’s a problem tbh

But they are not really spammers - it's legit listens, just several accounts connected to the same Spotify or whatever

Until we have artist/album stats and then we can see who’s gaming that, anyway

I think we should ask people why they did this. and to perhaps clean them up.

We don't allow multiple accounts in MB unless there's a clear reason (to run a bot or something) so technically these, by being the same people, are in breach of the MB rules

I guess for now we could get in touch with them and ask if they know / remember they have several accounts, maybe some are doing it by mistake

ah ha. ok, then this is pretty clear.

ima bout to link to that sp20 or whatever cluster mess of ratings mess to prove what is an issue (why)

I checked the CoC, I couldn’t find anything ;P But a cleanup wouldn’t be bad

but can't find the link

yes, do that please.

yea agree

ask

aerozol: well, these are effectively sockpuppets :) But I don't think any of them are done with a bad intent

I see people have issues with importing from lastfm or with other submitters, requiring them to remove all their listens and import afresh. That sounds like a good reason to have multiple accounts (one for clean imports from LFM for example, one for testing another importer)

agreed

So the question in the document was "what happens if we cannot get in touch"

I guess we can start with getting in touch and seeing from there

yp

So, I'll do that and bring this up again once I know why some people do it :D

Next is mayhem, with GSoC ideas pages!

yeah. one thing we could do is: tell us which account to keep, otherwise we'll delete the younger one.

right-o

I think we should allow people to export their listening history befor ewe take a desicion to remove an account

yvanzo: zas: outsidecontext: bitmap : akshaaatt: have you updated the ideas pages for your respective projects?

I don’t think we have anything in the MeB CoC, and the MB one jsut says don’t make multiple accounts for voting

(sorry for interruption mayhem)

no

please do that soon as we might be accepted and have students invading soon.

I updated the LB one, but there might be one more I want to add, but that can wait until the meeting is done.

so, please update those pages.

I want that last yeas

fin. back to reosarevok

ack

I want that last years bookbrainz idea with relationship ditor

outsidecontext: let's discuss about this tomorrow, the only idea added is the one for cover art processing, not sure if we want to add more at this point

zas: outsidecontext: Could autotag be a project for picard?

before we loose the ability for gsoc at all

Ok, seems like that's all for the meeting, so I'm going to finish it here, and people can keep talking GSoC ideas

A suggestion to help with the meeting... Perhaps when someone is absent without notice for three consecutive meetings, they could be dropped from the list of presenters (but could be added back if they ask)?

Agreed CatQuest it's going to be on the ideas page

monkey: <3

aerozol: launching is meant that if you download a playlist file from a web page, it will open a player for you.

mayhem: yes, likely

but the downloaded playlist cannot play local content since it has no idea where that is. or what it is.

mayhem, lucifer, aerozol: The new "Taste" tab (combining pins and love/hate feedback) is up on test.LB now: https://test.listenbrainz.org/user/mr_monkey/ta...

Ah, download playlsit from LB > play in your local system software. Cool, didn’t even think we might be looking at that

so the content resolver could be the app that launches when a JSPF file is opened and it then in turn opens the default app for handling m3u files.

rdswift: agreed

aerozol: yes, people with their own custom solutions want it. that was made clear at fosdem.

but my main motivation was to use it as the core for Funkwhale troi support.

rdswift: yes, that's kinda what I wanted to do, I just haven't been keeping track very well

and possibly later Navidrone (another funkwhile like project)

aerozol: could you be bothered to keep a small list of who was missing when making the recaps?

zas: let me know if you and outsidecontext want to discuss that idea further. lucifer and I plan to have the backend support for that before gsoc starts.

[musicbrainz-server] 14reosarevok opened pull request #2839 (03master…MBS-12891): MBS-12891 / MBS-7971: Show more info for inline artist/label search https://github.com/metabrainz/musicbrainz-serve...

reosarevok, that should be easy enough to get from the reports on the community forums.

reosarevok: good idea, will do

Thanks!!

mayhem: yes, I added it in Ideas page as a reminder, we need to complete the idea proposal though

Acknowledged, mayhem

mayhem: I have no idea how you would resolve that considering you don’t know where the local files are, so that sounds fun :D

akshaaatt: thanks.

zas: lets chat about it in the next few days.

mayhem: ok

aerozol: my resolver has a scan function and a DB. you scan your collection, and when you want to resolve it does a fuzzy resolve against your local collection.

reosarevok, perhaps the absense count could be kept on the wiki page?

this DB could also be useful for deduping a collection.

lucifer: you still about?

ooh

+1 rdswift

mayhem: ye

*yes

I had another ML idea for LB. I kinda originally scoffed at it when someone suggested that a company was doing that, but I have learned lots, so...

mayhem, #1 machine learning fan

mayhem, scan as in AcoustID scan or title/artist scan?

what if we trail an "ai" on the messy metadata we get (e.g. "blah blah blah - 2008 remaster") and show it what the cleaned up version is: "blah blah blah". we have loads of data.

and then, rather than a complete detuning, we just make the AI do it.

rdswift: file scan, then read tags, then save to sqlite.

aerozol: not sure what you mean of that, but what I really dislike is when its overhyped and full of BS.

but clearly there are some AI features that are noteworthy now.

mayhem: hmm, i guess that might work.

Okay, thanks. As a future project, an AcoustID scan would make deduping a bit more reliable (I suspect).

Just seen you making fun of daft “AI” ideas on here a lot mayhem, I know you’d be open to actual useful applications, don’t worry

I'm pretty certain it could work, but I am unsure if we could make it fit into production. scalability is an issue.

rdswift: deduping is not a key feature of this project.

resolving local content is, and that is what I am focusing on.

but my point is, that now we have a project with a DB, it could do deduping.

i mean yes it would work but i wonder how much training data we need to put into it so that it automatically identifies new cases.

Understood. A totally separate project idea (not necessarily MetaBrainz). It's just that we see so many questions about trying to dedupe using Picard.

lucifer: we have loads of data. :) I'm also not sure of it yet. which is why I am fishing for feedback if it should be gsoc or we throw it to the next students who wants to learn from ml.

rdswift: understood. I feel that deduping is just something picard should do, but it is clearly outside its wheelhouse.\

it seems separately sufficient so that new students can start work on it.

lucifer: agreed. maybe we just keep it to that.

if something pans out, turn into their gsoc project.

ok.

rdswift: and I wonder, if this is a tool for collections tagged with picard, one could make it pretty easy to find dups.

we have the data to make it easy.

rdswift: I’ve always felt there must be a simple solution for the Picard deduping. e.g. Add an option to check for files with the same name, and if so display a popup asking which one to keep (display some file metadata to compare, but mainly bitrate/format/size)

mayhem: Interesting idea. Certainly something to think about.

Potentially have an option to automate it with a bitrate/format preference list you can set

(for basic needs, assuming the files are tagged to the same MB entry and so on)

monkey: Hey!!

aerozol, I'm not sure how that would work. Unless I'm missing something (entirely possible), that would require Picard to maintain some sort of database (for file location), or find a way to use the user's "Collection" database somehow.

rdswift: this would be for when you hit ‘save’ - check target location then. Not a check of a whole DB

Hi jasje

hi akshaaatt

So I wanted to have your opinion on something

I was reviewing Rishav’s PR for the profile screen

rdswift: does that make more sense? Or is this all much trickier than I imagined

It seems pretty irrelevant to the work we are doing

akshaaatt: agreed

I am planning to drop the PR and ask him to close it