[Tango-L] Recognizing Tango Music

Shahrukh Merchant shahrukh at shahrukhmerchant.com
Mon Jun 7 13:50:45 EDT 2010


"What makes Tango music Tango music?"

Since there is a reasonable combination on Tango-L of musician types, 
computer science AI types, Tango DJ types and others who know Tango 
music in some depth, and of course dancers most of whom "recognize Tango 
music when they hear it," I thought I'd pose this question to the group.

More specifically, if one had to write an algorithm to identify Tango 
music, how might one go about it? You don't actually have to come up 
with or even propose an algorithm, since that's probably more along the 
lines of a Ph.D. thesis than a Tango-L posting :-), but what would be 
some essential objective characteristics that such an algorithm would 
look for?

Looking up the melody in a database of all accepted Tangos would no 
doubt be the simplest way (leaving aside the issue of inclusion or not 
of some fringe Tangos), but that's cheating and doesn't get to the real 
question.

We can even simplify the problem to exclude very "jazzy" Piazzolla-style 
Tangos, and to exclude electronic "techo-Tango." But I've heard some 
very Tango-esque Hungarian Tangos and Russian Tangos (not necessarily 
great for dancing), and presumably the algorithm would detect the 
Tangoness in these.

So let's say the algorithm puts out a score of 0-100 on how Tango-like 
the music is. The idea is that at the very least "obvious" Tangos should 
get a high score and "obvious" non-Tangos a low score.

It can't be based solely on superficial characteristics like the 
presence of a bandoneon (even though that could nudge the score up), 
since it would be too easy to fool the algorithm. Nor can it be based on 
things like the presence of the "chan-chan" at the end of a Tango (since 
most people will recognize even unfamiliar Tangos as such a few seconds 
into the song and don't have to wait for the end).

Any takers or thoughts?

Shahrukh



More information about the Tango-L mailing list