If today's hit songs all sound kind of the same to you, it might not just be that you're an old fuddy duddy who can't understand the stuff kids are listening to these days. University of Bristol researchers in the UK are presenting a paper this week that they say shows you can gauge the potential popularity of a song by running it through machine learning algorithms (shown here).
The research team naturally enough has focused on the UK top 40 singles chart (Little Mix's "Cannonball" is currently #1), looking back over 50 years. A website dubbed ScoreAHit explains their research in more detail, but in a nutshell they measure a song's hit potential through such features as tempo, time signature, song length and loudness, as well as harmonic simplicity. The researchers still have a ways to go in perfecting their system, though, claiming only a 60% accuracy rate for classifying a song as a Top 5 hit or not (that might be a relief to George Michael and Lady Antebellum, both of which released singles recently that the Songometer isn't giving much of a chance.
Among the findings are that more danceable songs now stand a better chance of going popular and that the late 70s/early 80s were highly creative musical times in that predicting hits was more difficult. The researchers say one important element of their system is that it accounts for changing tastes over time.
The results of the study differ from previous research, which has so far not been shown to predict hit potential. A possibly important qualitative difference with previous studies is the use of the time-shifting perceptron to account for evolving musical taste.
I'll sing for joy if you follow me on Google+