[mb-devel] Collaborative Filtering: Artist - Artist Relationships
(Summer of Code Proposal)
Frederic Da Vitoria
davitofrg at gmail.com
Thu Mar 22 09:40:24 UTC 2007
2007/3/22, Aaron Cooper <cooperaa at gmail.com>:
> On 3/22/07, sharon myrtle <sharon.myrtle at gmail.com> wrote:
> >
> >
> > Hi,
> >
> > I am Sharon Myrtle Paradesi, a senior undergraduate student from India. I'm
> > interested in participating in the Google Summer of Code '07 and the topic
> > 'Collaborative Filtering: Artist - Artist Relationships' of MetaBrainz
> > Foundation particularly interests me.
> >
> > Currently, I'm working on Collaborative Filtering Algorithm (specifically
> > tailored for Indian Movies) as my Final Year Project (nearly done). The data
> > we worked on had the form of user and movie details along with the ratings
> > the users gave the movies. We then calculated the correlation between the
> > users and recommended movies to new users.
> >
> > I have formed a basic idea of what the proposal should include.
> >
> > As far as data sources are concerned, these are the places where we could
> > search for relationships between two artists
> >
> > 1) Aritist Albums - two artists that are on a compilation together are
> > likely to be somewhat similar.
> >
> > 2) Search Logs - one user searching for a number of artists also gives
> > an indication of similarity.
> >
> > 3) Artist Subscription - the artists that MusicBrainz users subscribe
> > should also yield some information.
> >
> > /* the above mentioned are given in the proposal */
> >
> > 4) Genres – the artists whose albums belong to similar or same genre
> > tend to be musically similar.
> >
> > 5) Ratings/Comments of users – scanning the ratings of users to
> > various artists and/or comments of users to the albums also provides vital
> > information.
> >
> > .. etc
> >
> > I've got the following ideas to collect data and generate artist-artist data
> > –
> >
> > 1) Scan through all the data sources and write the collected
> > information into files (such as artist details, album details, user details
> > etc).
> >
> > 2) From these files, create and maintain tables of the various artist
> > names, and corresponding artists with whom he/she is closely related
> > (musically).
> >
> > 3) Now we can generate and analyze the relationships between the
> > artists.
> >
> > I'd be glad to hear any comments/suggestions on this. --
> > Regards,
> > Sharon.
>
> MusicBrainz does not have any information about genres (or custom
> tagging), ratings, comments (or discussions about artists/releases),
> and only has a very basic subscription tool right now. I am just
> commenting to say that I would love to be able to:
> * comment on artists/releases,
> * rate albums/songs,
I have been dreaming of this for a long time too. I believe this would
give new aspect to MB, which is currently purely technical.
--
Frederic Da Vitoria
More information about the MusicBrainz-devel
mailing list