WEBVTT

00:00.000 --> 00:09.840
I'm going to do a group presentation this is going to last eight minutes.

00:09.840 --> 00:16.320
I want to do a quick, quick, quick overview presents a new features that we've been working

00:16.320 --> 00:23.800
on and to present to you where we want to go and how actually to prepare some limitation

00:23.800 --> 00:31.800
to build music recommendations and music discovery features.

00:31.800 --> 00:39.360
So Fengwa is a nogeo streaming software, what you can listen to music, extracts, build

00:39.360 --> 00:51.040
playlists, actually our implementation of activity group support audio metadata, so we share

00:51.040 --> 00:58.800
the audio metadata between activity group, we created a bunch of new activity objects, so

00:58.800 --> 01:06.000
if you want to build a nogeo federation software, we reach out so we can share the same

01:06.000 --> 01:15.920
protocol like object definitions, we also use RSS and we have some centralized services that

01:15.920 --> 01:23.120
allows us to have a consistent database to avoid duplicates, because if two users

01:23.120 --> 01:28.000
upload the same track but in different files it's hard to be sure it's the same track,

01:28.000 --> 01:35.680
so we use music brain to have a unique ID to be sure they're actually speaking about the same

01:35.680 --> 01:44.240
music, we also use internet archive as a centralized service to get cover of hearts, so we've

01:44.320 --> 01:52.000
been funding by NLNet, we've been developing new features, mainly use the follows with

01:52.000 --> 02:00.480
listening playlists and favorite tracks activities, so these also are new activity group

02:00.480 --> 02:10.560
objects and as far as I know nobody else owns the federation use and we've been creating a link

02:10.560 --> 02:17.680
with listen brains, it's also a centralized service that allows to get your listenings and your

02:17.680 --> 02:24.480
favorite from other services like YouTube, but if I whatever sort of party services you want to use

02:24.480 --> 02:32.320
so we are like creating a bridge between governments and activity group if you want to

02:32.400 --> 02:43.200
we could probably have built a scrubler, so tools that record the listening activity

02:44.400 --> 02:50.240
as a decentralized services, but it was very interesting to have also listening in one place to

02:50.240 --> 02:57.760
build the recommendation systems because it requires to do statistical relationships between

02:57.840 --> 03:01.760
listening, so you need the more listenings you have, the more data you have, the more

03:01.760 --> 03:11.120
accurate the recommendations are, so this is where we want to build, build proper

03:11.120 --> 03:17.760
music recommendation and music discovery systems, so you have two ways to do that as I said,

03:17.760 --> 03:24.720
we have a lot of ways to do that, but mainly you can use user listings or you can use music

03:24.720 --> 03:36.560
metadata mainly general metadata, about general, we would like to use, we would pay the music

03:36.560 --> 03:44.560
brains to build a collective knowledge about types specifically general, general is very subjective

03:44.560 --> 03:51.680
so you need to debate to create them and also what would be very interesting is to have a general

03:51.680 --> 03:58.640
taxonomy, so for example if you have a general German based general you want to listen to all

03:58.640 --> 04:03.440
the self categories, so you need to know that I don't know the step is a subject of German based

04:03.440 --> 04:10.080
which is a subject of the and this is very hard to do and we didn't find any way to build this

04:11.280 --> 04:19.600
on using decentralized technology, so I guess using Wikipedia and this kind of collective tools

04:19.600 --> 04:30.640
it's also nice, what there's also other things we could do is we could pay the

04:30.640 --> 04:38.560
for example they have a list of artists that are considered as a fashion, so we could create

04:38.560 --> 04:47.520
ban list with that without having to do all the moderation work and the fact checking we could

04:47.520 --> 04:58.720
rely on as this kind of project, so for as I said we rely on centralized services it's a

04:58.720 --> 05:07.040
little bit strange for an activity pub project but we didn't find any solutions that was easily

05:07.120 --> 05:18.240
implementable, so if you have any idea feel free to reach out, also there's been some people

05:18.240 --> 05:27.280
willing to use think well to allow artists to get money and sadly think well it's not a music

05:27.280 --> 05:37.840
distributor, it could be nice to have this kind of institution because it could allow us to

05:37.840 --> 05:44.320
also have an income and a strong economic model but we have nobody on the team knowing how

05:44.320 --> 05:50.880
we project these music industry works, so about that also if you have knowledge or if you

05:50.880 --> 05:58.880
have ideas of how this could work, feel free to reach out, and of course I'll interrupt

05:58.880 --> 06:03.520
and to have a reliability is a political fight not only a protocol, sometimes we get lost and

06:04.320 --> 06:16.480
only implementing protocols and well we don't have the power of Spotify, so it's not a Spotify

06:16.480 --> 06:29.680
alternative yet and I don't think it will be okay thanks if you have any ideas about how to build

06:29.680 --> 06:37.360
decentralized recommendation systems and if you want to implement audio federation feel free to reach out

06:37.360 --> 06:45.360
we will probably work on an activity pub feature proposal for audio object next year and we will

06:45.360 --> 06:51.760
also probably try to work with a fitted test project to enhance federation with other projects

06:54.560 --> 06:56.880
so yeah that's it