Music Recommendation System for OCR

[Anorax]

Hey everybody, just wanted to update you on how the project is going. I have a basic recommendation system working now, which rates songs' similarity by how many tags they have in common. I've put the code on github, mostly just so you can look through the recommendation files if you're interested. Here is some sample output (first the similar song, then the tags they have in common).

For the song "A Different Kind of Peace":

Then the Healing Came orchestral,strings,cinematic

Shield of Legend orchestral,epic,cinematic

Justice for All orchestral,epic,cinematic

Captain of the Skies orchestral,epic,strings,cinematic

Enemy Underworld orchestral,epic,strings,cinematic

For the song "Adlehyde Castle Flow":

Insecta Robotica mellow,jazz,electric-piano

This Heart percussion,mellow,tempo-slow

Gemini Salsa percussion,jazz,tempo-slow

Holiday Frappe mellow,jazz,electric-piano

Fighting (7/8 Jazz Spiritual) mellow,jazz,electric-piano

There is still a real sparsity of tags--only 1,306 songs even have one tag, and only about half of those have more than one. My next move is to incorporate Remixer and source-game as tags, by extracting the meta-data from the MP3s. That should help, but it seems like to be successful the tag-based system needs more data. Once again, does anybody know how to see a list of all the tags in the system, not just that page with the top 70?

EDIT: forgot the github link... https://github.com/zygoth/OCR-Music-Recommendation-System/tree/master/

your best bet would be to contact Larry (liontamer) as far as accessing all tags.

As for updating all the tags for all remixes, I think that's part of the Patreon campaign, but I'm not sure? That one takes some time to do, obviously.

[zygoth]

It says re-tagged albums, not songs. Am I misunderstanding it...? As far as the songs go, wouldn't it be easiest to just open it up and let the whole community do the tagging?

[Anorax]

It says re-tagged albums, not songs. Am I misunderstanding it...?

whups, you're right. Damn.

maybe that should be a community project or something, suggest tags for the current library of remixes

[DarkeSword]

whups, you're right. Damn.

maybe that should be a community project or something, suggest tags for the current library of remixes

This is already in the works. We're still defining requirements.

[Nabeel Ansari]

It says re-tagged albums, not songs. Am I misunderstanding it...? As far as the songs go, wouldn't it be easiest to just open it up and let the whole community do the tagging?

One problem is that a lot of people aren't really qualified to make meaningful musical judgments like these tags. A community member may hear a saxophone and type "jazz", for example, even though his or her judgment is hasty and inaccurate. This is why Pandora selectively hires people to do so instead of letting everyone do it. Last.fm is the opposite, where the people do the tagging.

Still I have no idea what the OCR staff are up to and how they plan to update the system, but rest assured, it is coming, as staff members have said.

Edited December 2, 2014 by Neblix

[Anorax]

One problem is that a lot of people aren't really qualified to make meaningful musical judgments like these tags. A community member may hear a saxophone and type "jazz", for example, even though his or her judgment is hasty and inaccurate. This is why Pandora selectively hires people to do so instead of letting everyone do it. Last.fm is the opposite, where the people do the tagging.

Still I have no idea what the OCR staff are up to and how they plan to update the system, but rest assured, it is coming, as staff members have said.

If community tags were allowed, what if taggers were allowed to pick from a pool of tags (a "mood" pool, a "genre" pool [<--this could get messy], an "instrument" pool, etc.) instead of creating their own (because we don't need a "penis" tag for any reason). This way there's not a large collection of useless tags thrown about.

[DarkeSword]

That's what we're doing. Believe me, we're already way ahead of you on this.

[zygoth]

Cool. Do you have any idea yet of how long it's going to take?

[Liontamer]

Cool. Do you have any idea yet of how long it's going to take?

It's long-term, it's not something that'll be done right away. For the purposes of your project, IF you're going to rely on our currently populated tags, you may just wanna use the ones that are tagged so far as the pool of music you're analyzing.

[zygoth]

Hey all, just wanted to let you know what happened with this. I ended up doing both recommenders, one based on DSP and one on tags. You can read my final report and see all the recommendations here:

https://drive.google.com/folderview?id=0B6V741rUzUuvVkFMUXZoY19qTm8&usp=sharing

Let me know if you have any comments or questions, hope you enjoy!

[Nabeel Ansari]

That was a good project, there. I think your conclusions are spot on, especially that DSP advances need to be made before systems like this can improve.