Shared publicly  - 
Yllona Richardson's profile photoGary Stock's profile photoValdis Klētnieks's profile photoHeidi V's profile photo
This article reads very much like the early goals and objectives of the Music Genome project, which later became the basis for Pandora. I haven't seen the Pandora data set in many years. However, if memory serves, Pandora uses a checklist/questionnaire of 400 musical characteristics that are classified by human analysts (often musicians) against each song in their catalog.

The resulting analysis is then used as a collaborative filtering algorithm for "channels" and recommendations within Pandora.
Bloody hell.  That "research" ranks as one of the all-time most absurdly uninformed wastes of effort. Reading that article actually encourages ignorance. 

Seriously:  if a reader knows very little about music, reading that article would reduce their understanding of it.  It's the musical theory equivalent of a history lesson from Sarah Palin.
+James O'Connor -- yes, that pronouncement was quite a shocker!  I'm glad they said "accidentals" rather than "sharpy flatty things."  lolz

Their effort could improve -- perhaps far enough to produce a net knowledge gain for some readers -- in any of several simple ways. 

Two trivially obvious examples...

Possibly interesting: "I analyzed the chords KEY AND KEY CHANGES of 1300 popular songs for patterns."

Possibly valuable:  "I analyzed the chords of 1300 popular songs IN THE SAME KEY for patterns."

Both could still be misleading to the novice -- and flaky from any educated standpoint -- but less harmful to the reader.
Absolutely nothing here that should surprise anybody who's got even a basic understanding of music theory.
Add a comment...