Re: [mb-users] where to find ratings in db dump

On Fri, 2014-06-13 at 22:22 +0100, Marcel Ackermann wrote:
> Thank you very much Ian and Calvin!
> So I was looking for ratings of users of tracks (explicit feedback) as
> I am writing my masters thesis about music recommendation and the
> million song contest is only playcounts (implicit feedback), so I was
> hopeful when I read "user submitted annotations, tags and ratings" -
> but I understand that this is not part of what you release. 
> Are you aware of anything like this out there? (There is the yahoo kdd
> data but they don't include titles of anything, just IDs)
> Thank you very much,
> Marcel

Unfortunately, for this type of data, privacy is a really big concern.
Even with user data anonymized, given the song data (recording titles,
artist names, tags, ratings) it's sometimes possible to profile
particular individuals and find out who they are.

Your best bet might be to look for data sets where individual users are
explicitly making their listening habits and rated/tagged tracks public.

A few places to consider:
* It might be worth contacting the folks at http://libre.fm and asking
them if there's a possibility of getting a dump of public user data for
your project.

* Last.fm also has a lot of this sort of data, but you'd be limited to
extracting it via their API, e.g.
http://www.last.fm/api/show/user.getPersonalTags - which will be subject
to rate limits (5 req/s).

* That said, we do at MusicBrainz have user preferences for "Allow other
users to see my subscriptions / tags / ratings" - I wonder if we could
filter data based on this preference to only include public information,
and make that available.

-- 
Calvin Walton <[email protected]>