Incentivized sheet music transcription platform

ldwm · April 25, 2020, 9:12am

Description:
A platform that incentivizes musicians to transcribe all music ever recorded. Like guitar tabs websites, musicians collaborate on song transcriptions using a web score editor (like Noteflight, Flat (dot) io).

Every edit to the score is saved so every revision can be retrieved and branched from (think Git/Irmin). Collaborators may mark sections of the score and create tickets/issue to document that certain parts of the song are missing, incomplete or erroneous (like Github issues).

The score may be submitted to a committee (Token Curated Registry) for review. If it gets accepted, the score gets assigned to the main unique Link/Anchor that belongs to the MusicBrainz ID or ISRC for that artist/album/recording.

Every user who contributed notes and edits to an accepted score is awarded with a “stake” in that song relative to their contribution. When researchers, academics and teaching institutes use the API to query the data (“get all songs after 1800 that contain a C Major chord with a violin instrument and a tempo >+ 180 etc”), the transaction fee is divided over the users that hold stakes in the scores that are included in the query results. “Stakes” have a floating value and may be traded.

Resources: See https://parture.org

Similar app or site: Project Gutenberg, Google Books, OpenScore

Team Size: 2

I am at…: Fundraising and technical design. I had been considering Solana, CODA and Holochain and think HC is the right tool due the native git-like object blob storage (DHT) and scalability promises. Solana has limited storage and CODA does not have programmability yet.

Skills: list any skills that are needed for your project. Includes technical (languages, technologies) and not-so-technical (fund raising, community outreach). Rate your team proficiency from 0 - 10 for each skill

Holochain knowledge: (4)
Front-End: Elm (8)
Back-End: Rust(5), Crystal (7)
UI/UX: CSS(9), Figma (7)
Others: fundraising(8), music (8), business development/sales (3), Token Economics (7), State Machine modeling (6)

Commitment: my life since 2008. 4h/day during Corona.

What do I need: what do you need for your project to go further?

Modules on:

Currency
Reputation (non-transferrable)
interest yielding Stake/Stock (transferrable)
discrete/continuous Committee Voting

More Holochain knowledge about:

the performance of walking and validating very long chains (imagine every note in a song being an entry)
communication between ephemeral and persistent DHT DNA (material from @pauldaoust)
security: how to influence which entries nodes replicate and which not (think overflow attacks)
DHT indexing

Technical team members not afraid to change the world

hedayat · April 25, 2020, 10:43am

@ldwm this idea is beautiful I really love it. I also think so that Holochain is the right tool to implement this.

Can you explain these 2 more in detail?

that is true but we need to obstacle one issue here which is the file size. @ogu83 challenging this issue and we are going to discuss it soon in Hackalong

ldwm · April 25, 2020, 2:00pm

Great to hear!

Transferrable interest yielding Stock in scores
The second is the easiest to start with. Imagine that an index (“Registry”) of all artists, albums and recordings (like from MusicBrainz or Spotify) is maintained in the network. It is maintained by participants using votes, like a Token Curated Registry from cryptocurrency spheres.

Suppose we collaboratively work on a score for Led Zeppelin and we both contribute 50% of the notes. We then submit the score to the Registry for acceptance. Participants vote on whether the score we made matches the song (“Stairway to Heaven”) in the index, and whether it is an improvement on a previous score linked to it (if any). If the vote succeeds, the score we made is linked as the master (think Git HEAD) score for the song. This means its linked to the root index anchor, used by the search engine. The acceptance rewards us with Stock in the song we transcribed. In this case, you get 50% of its Stock and I get 50%.

Then, imagine thousands of people have done this and there is quite a considerable database of reliable and complete scores. This would make it interesting for AI researchers, academia, composers and others to perform queries on the complete dataset (this requires the DHT indexes). For this, a zome accepts a query and a transaction fee. The query could be a composer looking for all Led Zeppelin songs that contain more than 200 measures. The query is performed and results in a number of scores linked to the Registry. Suppose that the score we worked on is the only query result. The transaction fee is then divided over all transcribers who contributed to the score linked to the official Registry entry. Since we both hold 50% Stock in the song, we are both awarded with 50% of the transaction fee. This is what I meant with “interest yielding stock”.

This Stock is transferrable, meaning we could gift it to another user, trade it or sell it. For example, suppose we are in love with the worst song Led Zeppelin has ever made. We both transcribe it for 50%. But since we do not anticipate it would come up in any query in the future (rendering us interest), we decide to sell the Stock we earned (to other users in the market).

Non-transferrable reputation
Because we both contributed to completing the Registry and got our score accepted, we are awarded Reputation. This Reputation is used (weighted with amount of staked currency) when casting votes ourselves for the Registry. The Reputation is tied to a end user account (not agent). Reputatation works on Stack Overflow. For example, users with certain amounts of reputation are able to merge and close issues in scores. Reputation also allows a global musician score. If you need a song transcribed for a complex guitar genre, you can lookup which user has the best reputation for that genre with the best chances of being the most accurate transcriber, and invite them.

guillemcordoba · April 26, 2020, 4:45pm

I also like the idea a lot. The git like functionality can be implemented with _Prtcl, a project I’ve been working on that wants to provide git like functionality to things other than code.

What are your thoughts about membranes? Who should be allowed to enter the network. Is there a big open DHT with everyone allowed, or many small interoperable communities in which you can enter if you have certain reputation?

ldwm · May 27, 2020, 7:31pm

Thanks for your comment. No, it’s a big free-for-all. Everyone is allowed to enter the network, but some capabilities are reputation-dependent like StackOverflow. For example, editing artist names (like fixing spelling mistakes) in the main metadata registry of songs and artists should not be possible until a certain trust level is reached.

ldwm · May 27, 2020, 7:38pm

Also, thank you for your project. At first a few weeks ago i didn’t understand its importance, but now I see that your project describes exactly this multi-perspective-single-cas system I need and have been trying to get my head around for a few years.