#metabrainz

just go ahead and ask your questions here, better than PM

ok! So I'm wondering if I could get some input or advice on a pre-proposal for a GSoC application. I'd like to contribute to discoverability of new music through the use of semantics in written content and visual content relating to albums. In essence what I'm getting at, is looking at content-based recommender systems from another angle that's based on natural language processing and encourages serendipity in the process of

discovering new music, that leaves the user with a sense of autonomy and control as opposed to the AI DJ's that have taken over spotify. I believe that with the metadata that's already present in MusicBrainz database, along with a user's taste which is imported from Last.fm or an external source, a semantic recommender system can serve as an assistant to a user searching for new music to listen to.

first off, do you know: https://listenbrainz.org -- another project of ours, like last.fm, but open source.

and we have recommendations already: https://listenbrainz.org/user/rob/recommendations/

these are based off collaborative filtering -- so far we have nothing semantic.

this sounds very interesting, but also very ambitious. I hope you already have some great tools/data to bring to the project...

<mayhem[m]> "first off, do you know: https..." <- Yes! I came across it as I was scraping lastfm data for a personal project of mine :)

<mayhem[m]> "and we have recommendations..." <- So I was initially searching for a way to scrape semantic data off of a website like RYM so I could have a dataset that had mood-based data in relation to an album object. A professor at my university suggested I look into recommender systems with serendipitous results, specifically for music information retrieval (we both enjoy listening to different kinds of music).

My major areas of research in my academics is in data/information visualizations + social computing, and recommender systems overlap with both of these areas.

MusicBrainz actually has mood support in the DB, but we haven't actually gotten around to building the mood UI or deciding how to represent moods.

mayhem[m]: OOOHHHHHH

🤌 well then... Count me in I suppose. I have some ideas I'll sketch tomorrow for how that can be factored into visualizations + recommendations. I'm part of a HCI group at my Uni so i'll def ask them for input too!

question: are you here for summer of code or as part of your existing uni work?

Well I don't have anything planned for this summer as I'm graduating in May and starting grad school in september, so I figured that GSoC might be worthwhile!

got it. well I look forward to what you come up with!

ok, off to bed, 2am here. nn

bitmap zas mayhem : instantiated accounts for y'all in SSO candidate tool #1 (kanidm)! with the way the initial credential setup (passkey or password+totp) works, i have to generate a unique link for you that expires an hour after creation, so **sometime after 10am EDT (UTC-4) on sat 3/15 or sun 3/16**, please ping me so that i can get you a link. thanks again for being willing to test SSO tools!

i don't have service providers (applications) set up just yet, that's to come in the near future :)