| Who | When |
Messages | |
|
|
|
| |
Messages 12-11 deleted by topic administrator 07-21-2006 08:56 AM |
Thomas Terashima
|
10
|
 |
|
01-22-2003 07:48 PM ET (US)
|
|
The final step would be to link the recognition of "new" melodies to a peer-to-peer database, a "Napsterized RIAA". Think of it as "Melancholy Elepants on crack (tm)".
tom -=W=-
|
aha
|
9
|
 |
|
01-22-2003 05:06 PM ET (US)
|
|
Edited by author 01-22-2003 05:29 PM
Eli: another way of looking at it: 22,000,000 - blues = 60,000 XAWWT(not that there's anything wrong with that)
|
Songdog
|
8
|
 |
|
01-22-2003 04:33 PM ET (US)
|
|
Edited by author 01-22-2003 04:33 PM
Eli - Parson's technique was intended for classical themes, which is a significantly restricted set compared to the set of all recorded media from the last fifty years or so. Of course, if you stretch Parson's same-up-down to twenty digits you've got about three and a half billion melodies covered. Another alternative (that taken in Barlow and Morgenstern's A Dictionary of Musical Themes is to represent the initial notes of the melody, transposed to C. That gives you base 7 for strictly diatonic melodies, so ten diatonic notes can identify more than 280 million unique melodies (ten chromatic notes would get you almost 62 billion unique melodies).
|
Eli the Bearded
|
7
|
 |
|
01-22-2003 04:24 PM ET (US)
|
|
Edited by author 01-22-2003 04:25 PM
Same, up, down, huh? That's three positions, ten digits of that is just a tad under 60,000. Not exactly a huge universe. Consider that CDDB claims to have "22,377,077 Songs" in their database. Many of those will be repeats of course, but that many?
|
| rananite
|
6
|
 |
|
01-22-2003 04:01 PM ET (US)
|
|
Quite a nice tool, but I can't help but think it will be abused by music companies to blackmail independent artists by threatening infringement lawsuits (especially now that copyrights can last 90+ years!).
Plus, once searches like this become so easy, wouldn't companies require composers to perform internet searches on their new tunes before they are published, to avoid liability? Given that there are always superficial similarities between different kinds of tunes, how would that stifle creativity?
Scenario 1: "Dear Mr. Smith: It has come to our attention that when every other note in the bass line of your recent composition, 'Wireless Sonata in 802.11g' (2003), is hummed into our proprietary computer search algorithm, one of our client's works, 'Telephone Hold Music Selection #712 (light jazz)' (1986), appears in the search results. Clearly, your work is an infringement of our client's copyright. Please send $500,000 plus all past and future royalties from your work or we will sue your sorry ass into the ground."
Scenario 2: "We commissioned the new piece from you 8 months ago. Why don't you have anything yet?"
"Well, I've written 14 different songs since then, but every one of them got at least a dozen hits on Google Tunes, so your corporate policy said you couldn't accept them..."
|
| cypherpunks
|
5
|
 |
|
01-22-2003 03:29 PM ET (US)
|
|
I hear that humming was responsible for a 4% decline in CD sales last year. Watch out.
|
aha
|
4
|
 |
|
01-22-2003 02:32 PM ET (US)
|
|
Edited by author 01-23-2003 04:06 PM
Cool! Thanks, Songdog! Parson's code, from "Directory of Tunes and Musical Themes" (Spencer Brown, 1975) by D. Parsons.
|
Songdog
|
3
|
 |
|
01-22-2003 02:22 PM ET (US)
|
|
Yes, aha, see my post here for links and references :)
|
| aha
|
2
|
 |
|
01-22-2003 01:40 PM ET (US)
|
|
Edited by author 01-22-2003 01:51 PM
Quite a while ago now I heard of another, more elegant technique, based simply on whether the subsequent note goes up, down, or stays the same as the previous note. After a sequence of five to ten, you can nail any song title. It might have been a book. You don't need to sing well, or sing at all. Anyone remember this? First same same same same same same up down up up = Jingle Bells.
|
| Ben
|
1
|
 |
|
01-22-2003 01:27 PM ET (US)
|
|
|