Not sure if I need to do anything more, or if things just get batched up / looked over at certain times etc :)
2019-04-01 09147, 2019
bluefantail
Read most of the contribution guidelines and think I've mostly been on track - hopefully anyway (let me know if not)!
2019-04-01 09130, 2019
Wizzup has quit
2019-04-01 09112, 2019
Wizzup joined the channel
2019-04-01 09109, 2019
ruaok
bluefantail: it not you, its us.
2019-04-01 09122, 2019
ruaok
we've been working so hard on the LB features recently we're all doing something else for a bit.
2019-04-01 09130, 2019
ruaok
that said...
2019-04-01 09134, 2019
ruaok prods iliekcomputers
2019-04-01 09141, 2019
bluefantail
Haha - no problem!
2019-04-01 09144, 2019
bluefantail
Just checking in
2019-04-01 09112, 2019
bluefantail
I mean if you ever need a hand with anything ..
2019-04-01 09137, 2019
ruaok
are you generally interested in helping with LB?
2019-04-01 09138, 2019
bluefantail
I'm free occasionally around work :P
2019-04-01 09145, 2019
ruaok
we're always keen to get more help.
2019-04-01 09119, 2019
Mr_Monkey
New article posted on the blog
2019-04-01 09133, 2019
bluefantail
I've been interested in the whole metabrainz thing from a distance for a while - just been keeping an eye on things
2019-04-01 09151, 2019
bluefantail
trying to understand goals better etc
2019-04-01 09103, 2019
ruaok
Mr_Monkey: already shared on my FB. I'm happy that we finally got *THAT* releasesed. so much work!
2019-04-01 09107, 2019
bluefantail
I think it's pretty cool / we're severly lacking projects like this
2019-04-01 09142, 2019
ruaok
bluefantail: like most open source projects, our goal is total world domination. what else is there? :-p
2019-04-01 09149, 2019
bluefantail
And this is the only one I know that looks like it could really bring good to the whole music discovery / and open media DB sort of landscape
2019-04-01 09103, 2019
ruaok
yes, agreed.
2019-04-01 09116, 2019
ruaok
the goal that I am keen on is to get a recommendation engine working.
2019-04-01 09132, 2019
bluefantail
So you might have seen my poking at a few things on the discourse just trying to understand more :)
2019-04-01 09132, 2019
ruaok
I feel that a lot of recommendation engines are just shilling to sell more major label content.
2019-04-01 09154, 2019
ruaok
and there are no open source recommendation engines -- our goal is to change that.
2019-04-01 09119, 2019
bluefantail
And yeah I kinda got interested in the LB API first, and then was slightly disapointed when I couldn't query the API from a browser XD
2019-04-01 09150, 2019
pristine__
ruaok: hey. So instead of saving and loading the trained model, I directly passed it to recommend.py and guess what it worked. Recommendations are being generated for the users. So we know that the problem is in saving the model.
Spark just blows my mind away. lol. would have to figure out why it isn't getting saved
2019-04-01 09106, 2019
bluefantail
ruaok: Haha - yeah fair enough. Yeah I gave my data to last.fm and google for years, and only in the last few did I realise it'd never occured to me what if I could own that data and ship it around to get reccomendations out of different services, or indeed if everyone's could be used to improve everyone's recomendations generally.
2019-04-01 09130, 2019
ruaok
yep, all that.
2019-04-01 09115, 2019
reosarevok
Freso, ruaok: can you think of a good tweet for our newest MB development we just blogged about?
2019-04-01 09125, 2019
reosarevok has slept like 3 hours and his brain is not working
2019-04-01 09147, 2019
ruaok
I can have a look when I pull my head out of accounting
2019-04-01 09113, 2019
aidanlw17 joined the channel
2019-04-01 09120, 2019
pristine__
ruaok: there is one user for which recommendations is an empty list. We should have top hits or something to recommend to such users
2019-04-01 09120, 2019
ruaok
do you have a recommendation for me?? I'd love to see that.
(huge dimension PDF, zoom in and have a look around)
2019-04-01 09140, 2019
alastairp
it represents the "similarity" of genre labels from different places, based on the cooccurrence of these labels being applied to the same MB recording
2019-04-01 09107, 2019
alastairp
e.g. if all recordings that have label x from lastfm also have label y from discogs, we say that x and y are very similar
2019-04-01 09131, 2019
alastairp
there are some really great clusters being built up here
2019-04-01 09146, 2019
aidanlw17
alastairp: that's interesting! I'm working on my gsoc proposal which is related to some of the recording similarity tasks with annoy
2019-04-01 09106, 2019
aidanlw17
do you think that the clusters of genre labels could be incorporated into the distance calculations for similarity?
2019-04-01 09148, 2019
Mr_Monkey
That's really good alastairp! There's some interesting clustering too like new-age being quite close to opera, and modern jazz clumped with soundtrack & ost.
2019-04-01 09102, 2019
Mr_Monkey thinks of https://www.youtube.com/watch?v=n2rVnRwW0h8
2019-04-01 09139, 2019
alastairp
Mr_Monkey: the next thing we should do with this data is actually look at the examples which are causing these clusters to form
2019-04-01 09103, 2019
alastairp
is it because of folksonomy tags (last.fm, tagtraum), which therefore means that people don't know what they're talking about?
2019-04-01 09140, 2019
alastairp
or is it because we don't have many example recordings tagged with those items, causing the few annotations that we have to build clusters when they don't really make sense?
2019-04-01 09157, 2019
ruaok
hmm. paying people on april 1. I should pay everyone in INR today, I think. except our indian team members. they get paid in EUR. yeah. :) :)
2019-04-01 09132, 2019
alastairp
aidanlw17: yeah, in the future something like this could be used for similarity, however the similarity task that we want to work on for SoC should be focused only on content-based similarity for now
2019-04-01 09106, 2019
Freso
reosarevok: "We’re about to launch our most ambitious change to our editing system since basically forever! Read the blog post to learn what’s happening! https://link-to-blog #opensource #ai" ?
2019-04-01 09133, 2019
alastairp
Mr_Monkey: although, check out the graph a bit more carefully.... opera/new age are next to each other, but if you back-track the lines then they're a part of two distinct clusters
2019-04-01 09146, 2019
alastairp
the vertical lines mark the similarity
2019-04-01 09106, 2019
alastairp
perhaps I should look into seeing if we can draw a line between labels that are part of separate clusters
2019-04-01 09153, 2019
aliamo joined the channel
2019-04-01 09103, 2019
Mr_Monkey
Well, I suppose I was talking about the similarity between the new-age cluster and the classical cluster
2019-04-01 09137, 2019
Freso
reosarevok, ruaok: Also, re: Mapillary: https://www.mapillary.com/app/user/freso :x – the app works fairly well, but is a huge battery drainer (as would be expected)
2019-04-01 09104, 2019
alastairp
yeah. I suspect that could be down to people naming everything that has violins in it as "classical" too, though :-P
could you do me a favor and write a script that takes this info and then cross references it with messybrainz and outputs nice readable artist/recording names and ids?
2019-04-01 09123, 2019
ruaok
I have no idea! I've not been able to look them up!
2019-04-01 09145, 2019
pristine__
ruaok: Okay. I will :)
2019-04-01 09122, 2019
yokel has quit
2019-04-01 09148, 2019
aliamo has quit
2019-04-01 09136, 2019
baniket has quit
2019-04-01 09103, 2019
yokel joined the channel
2019-04-01 09146, 2019
aliamo joined the channel
2019-04-01 09147, 2019
BrainzGit joined the channel
2019-04-01 09108, 2019
aliamo has quit
2019-04-01 09144, 2019
pristine__
ruaok: do we have MBIDs for every song?
2019-04-01 09122, 2019
ruaok
no. far from it. :(
2019-04-01 09127, 2019
pristine__
And what does cross reference exactly means?
2019-04-01 09155, 2019
ruaok
don't print MBIDs, but print human readable text.
2019-04-01 09108, 2019
ruaok
or at least, first humans, then MSIDs/MBIDs
2019-04-01 09155, 2019
pristine__
Then i can just modify this script to print human readable text.
2019-04-01 09111, 2019
ruaok
anything to make it easier to understand the output. :)
2019-04-01 09118, 2019
pristine__
And how are we actually going to build playlist? I mean the user can actually click on 'play' and listen to the song. Like we have done for spotify.
2019-04-01 09102, 2019
pristine__
We won't be recommending just names, right?
2019-04-01 09121, 2019
ruaok
that's a very good question. I don't think we have an answer just yet.
2019-04-01 09136, 2019
ruaok
perhaps we should limit the recommendations to tracks that we can play in spotify.
2019-04-01 09149, 2019
ruaok
but these are things for you to solve in your GSoC project. :)
2019-04-01 09114, 2019
pristine__
Okay :)
2019-04-01 09122, 2019
baniket joined the channel
2019-04-01 09146, 2019
baniket
Mr_Monkey: hey !! I asked some questions on the irc yesterday, it'd be great if you can please answer them :)
2019-04-01 09106, 2019
Mr_Monkey
Hi baniket! I was actually having a look at it :)
2019-04-01 09145, 2019
Mr_Monkey
> there is no initEditorRoutes() function but there is a function like this for every other route, is there any reason behind this
