Bruce Warila | Fri, October 10, 2008 at 12:07 PM This is not a product announcement. Heads up, I have something like Pandora in a box.
Note: Pandora uses humans to analyze songs. We use unbiased computers running refined algorithms. THIS PRODUCT IS NOT AFFILIATED WITH, OR POWERED BY PANDORA (the music recommendation site for consumers). Think metaphor here...
If you are building products or services for artists or music consumers, here's a music recommendation product we (Platinum Blue Music Intelligence - PBMI) have been working on.
PBMI has created a virtual appliance (other product variations will be forthcoming) that enables any site to rapidly install software that does the following:
- SongSeeker (the name of the product) generates accurate, computer-based, sounds alike listening recommendations.
- SongSeeker process any number of songs.
- Into rows of songs (MySQL).
- Where every row contains N number of songs.
- Where every song in the row is sorted by sonic relevancy (close to far, numeric measurements provided).
- This product scales to infinite.
The underlying recommendation engine was created by the Centre For Digital Music in London.
Note to the press: The perceived problems with machine-based recommendations are probably due to poorly designed interfaces and implementations; the problems may have nothing to do with the ability of the computer/algorithm to make quality sounds-alike recommendations.
Note to potential customers: How you implement a sounds-alike recommendation product is as important as picking one. Creating graphical representations of the output that perhaps color-codes recommendations by genre or by other meta identifiers could be the difference between a poor end-user experience and a great one. SongSeeker enables developers to rapidly create propriety interfaces.
Note to artists: This is the type of product that will get your songs found effortlessly. When I write about the Celestial Funnel, this is some of the technology (the dawn of the technology) I am referring to.
Possible use cases:
- Recommend/find similar sounding songs to/for music consumers.
- Recommend/find similar sounding artists to/for music consumers.
- Matchmaking for artists (artists that have sonic synergy).
- Sort songs/artists by sounds-alike features.
We are competitively pricing this product to fit the needs of every business; regardless of size.
Contact me directly for more information.
View Printer Friendly Version
Email Article to Friend
Reader Comments (2)
Have you forgot about lyrics?
Where's your algorithm that will funnel songs i like based on lyrical content?
Your software is clearly not designed by songwriters.
"Your software is clearly not designed by songwriters."
Mike - Thank god for that!
All kidding aside, please note that I stated that to create the best end-user experience, other meta identifiers should be used. Songs that are adequately tagged (by the creator, socially, by experts, or by machine) can be sorted/found/extracted by lyrical content. If a customer (one of our potential customers for example) is deploying a sounds-alike, listening recommendation product, we recommend that they empower artists to tag their songs. SongSeeker + tags solves the problem you describe. Although, I don't think it's necessary to have tags to introduce people to new needles within a multi-million needle haystack. Moreover, it's pretty simple to index lyrics and to then provide some basic search / extraction features to enhance the results we produce.
Mike - Thanks for the comment. I put up these posts to learn about the objections people have.