Musicoin Music Discovery, Categorization, and Tagging Systems' Design

Hey all – this was pinned in Musicoin’s discord site for a while but I’m moving this project over here since this forum seems like the best place to have these discussions from here on out.

Basically the gist of this project is that I’m interested in having discussions about how Musicoin’s “Search for Tracks” function can be improved so that it provides a better experience for both musicians and listeners. As most people here probably already know, music discovery systems on most platforms right now aren’t very good because it tends to reward incumbents unfairly (usually in collusion between major labels + the platform), are non-transparent, not to mention very disorganized in general. As a result, I don’t think anyone really trusts music-tech platforms to really be arbiters for what they consider to be “quality content” since they have to more or less have to do that discovery work on their own. Does it have to be that way, though?

https://docs.google.com/document/d/1k7RL1uKHFiU6T75jD0zsqPYXfN0cI5cfKCP3p1w5Jq0/edit

I wrote this PRD (product requirement document) as a start but after mulling it over a bit I realized this project was much bigger than originally thought, so I tried to focus the scope to just the Category and Tagging systems for now. On a very basic level, how do we organize our tracks on here by genre, artist type, geography, use-cases in a very broad sense? If you make it too broad you end up losing interesting content due to over-centralization, if you make it too specific (e.g. dark ambient dub-step with reggae influences made by your 2nd cousins) you get the same due to increased noise and confusion by the listeners. It’s a difficult balance to achieve but I do think that if we put some real thought into it we could probably come up with something better than exists anywhere else right now.

My tendency is to favor simplicity and clarity over everything else (for transparency and reduced maintenance loads) so that will be my agenda from my end. 🙂

I’m a long-time musician here – I have classical training and a few years experience studying musicology at the Ph.D level but my style is improvised ambient music with a cross-genre approach. I’ll make edits and revisions to the document as time goes on (very much WIP) so let me know what you think!

Music Discovery System

Take the chart above with a grain of salt, btw. Just something I whipped up real quick.

last edited by Ryan Tanaka

Every play by every listener on every song is stored on the blockchain, a recommendation engine based on scraping all the plays a song has had and searching for which other songs have had the most plays by those listeners would be useful alongside genre tagging. It would work on the artist level too. Songs most tipped by listeners who tipped this song, artists most played by listeners who like this artist, that sort of thing.

administrators

@ryan-tanaka good ideas. we should held an online discussion if you love to organize.

@im I’m hoping this thread will be the main source of discussion for this topic – it’s a huge undertaking but I think the project would benefit from a clean categorization scheme, even if it’s not perfect. The only useful data is clean data, after all. 🙂

Food for thought: Pandora’s Music Genome system is proprietary but you might be able to reverse engineer some of it based on what we know. Page 7 of this slide has some of the tags that are embedded in some of the tracks.

https://docs.google.com/presentation/d/1w9B2qzFHebLjz11ZGQVWTgIZvSOBzftGuOFg0eQ3Ns8/edit#slide=id.g11cc9ae2eb_0_86

Pandora hires people with music theory and musicology backgrounds to tag the tracks manually (a little surprising considering how goofy some of them are) but I think that these tasks could be built into a blockchain protocol as potential rewards for keeping the tags fairly clean. It would also remove the burden of self-tagging done by musicians – which, at least in my experience, tend to be more of a chore than anything else.

I think that music genres (Pop, Rock, Classical etc.) should probably be broken down in a generic and neutral way for ease of maintenance – and possibly even restricting each track to one per genre at first to reduce the noise and confusion that exists right now. Best not get involved in categorization battles here. But tagging should be done in a decentralized way, imo.

last edited by Ryan Tanaka

Looks like your connection to Musicoin Forum was lost, please wait while we try to reconnect.