I’m interested in selfhosting, if possible, an equivelant to last.fm – it would analyze the history of what I listen to, and provide me with recommendations, and listening history reports.
Aside, last.fm as a federated service would be quite interesting. It would be neat to add a federated social media aspect to it.
How would these recommendations be generated. Will you host for loads of people or just sit there with your single datapoint?
That’s a fair point that I had not considered – it’s a shortcoming in the premise of my inquiry. I wonder if it’s possible, if at all, to create any recommendation service that doesn’t compromise on user privacy. It may not be, as it would require a user’s history, which, given enough entries, can be used to identify them.
Its possible with anonymous data. But you do need other user’s data. So it would need to somehow be not just self-hostee but also federated. With lots of dsts from other servers
The issue, I think, is that having access to a user’s entire listening history could very well be used to identify that user – one’s full listening history is likely to be rather unique.