[Rhythmbox-devel] Exponential playlisting - a novel querying model

Rhythmbox, as the majority of music management systems, still rests on a
linear model of playlisting - I choose a set of terms which narrow my
selection from all available tracks until I have the subset that are
restricted by my query terms.

This model is powerful, yet fails to cope with the relational
associative methods we use to describe music, in particular the very
ill-defined properties of mood, style and genre.

Nonetheless, there exists an existing mapping scheme of related music
that remains unexploited by (I believe) all databasing systems, and one
that I believe Rhythmbox, due to the combination of interface and
querying engine, is ideally suited to employ: exponential playlisting
using compilation albums as reference points.

This functionality addresses the case where one is trying to create a
playlist of 'related' songs, and is of utility to those with large
collections. Consider the typical music collection. One has a number of
single artist albums, and a number of compilation albums. If one wants
to create a playlist for 'similar' music, one must select all the
artists that play that style of music, and then all the compilation
albums similarly, and enqueue each to the playlist individually.
However, this fails when there are either a large number of artists and
albums (time consuming) or if one forgets the existance of a given
artist or album (easy to do in a 10000+ song collection). Genre-field
mapping is ill-defined and near useless in these situations.

A solution would be functionality via which one could query on an
artist, retrieve the albums that they feature in, then clear the artist
search query while querying on the albums returned to retrieve a list of
artists appearing in those albums. Query on these artists, while
clearing the album query. Now we have a set of albums, including
compilations, which any of these artists appear on. Coexistance on a
compilation album 'links' artists together. Repeat, until the playlist
is sufficiently well-populated (or the number of records returned
remains static). Now all related tracks (by album-coexistance of the
artists) are in the playlist. By this method, all your 'similar' music
is in in the playlist. If some compilation albums (such as 'samplers'
from a label) bring together clashing records, then these could be
excluded in the next stage by a simple subtractive stage.

An initial semi-automatic implementation could prevent updating of the
search in the browser (a 'Hold Query' button) while the search critera
are changed. The sequence of events as carried out by the user would
then be:

1. Select artist. The Album and Track queries update

2. Hold Query on.

3. Deselect Artist in the Artist box. Select all albums in the Album box.

4. Hold Query off. Artist and Track boxes update with artists and tracks
on the selected albums.

5. Hold Query on

6. Deselect albums, select artists shown as on these albums.

7. Hold Query off. Album and Track boxes update with albums and tracks
listed with these artists.

8. If the list of tracks is satisfactory, enqueue listed tracks.
Otherwise return to step 2 and repeat.

At each step of the above, inappropriate artists or tracks could be
excluded by deselecting while the 'Hold Query' is on. One 'Hold Query'
button would not clutter the interface, there may be an even more
intuitive method for allowing the user to control this functionality.

Alternatively, another section of the interface could perhaps be more
appropriate for this system to exist within, and the system could easily
be extended to be more powerful and complex. Nevertheless, the simple
model proposed above would adequately address this problem.

I suggest this as a proposal and as a challenge-to-implementation, and
as a catalyst to stimulate sicussion of truly novel playlisting concepts.

best regards,

Mark A G Jones
Nanomaterials for Quantum Information Processing
Department of Materials, University of Oxford,
Parks Road, Oxford OX1 3PH, UK

tel: +44-1865-273700    tel direct: +44-1865-283342
fax: +44-1865-273789    E-mail:     mark jones materials ox ac uk

Group Website:		http://www.nanotech.org
Collaboration Website: 	http://www.qipirc.org
Department Website: 	http://www.materials.ox.ac.uk

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]