| Author
|
Some interesting reading material
|
TiMMY
IsraTrance Full Member
Started Topics :
100
Posts :
1480
Posted : Jun 18, 2006 11:26
|
|
Trip-
IsraTrance Team
Started Topics :
101
Posts :
3239
Posted : Jun 18, 2006 11:31
|
so who decides what is good? the program was programmed to watch parameters anyway...
"U2 hits" - pfft... if that's what the future of music holds - it's not really surprising.
  Crackling universes dive into their own neverending crackle...
AgalactiA |
|
|
|
l337
IsraTrance Full Member
Started Topics :
55
Posts :
817
Posted : Jun 18, 2006 13:10
|
|
old |
|
|
|
l337
IsraTrance Full Member
Started Topics :
55
Posts :
817
Posted : Jun 18, 2006 13:13
|
However tho i think it is a load of crap in the sense that....why are us humans now relying on a computer to go as far as telling us what will be a hit....
if it will be a hit one should know it will be a hit.....not have to rely on a computer program |
|
|
|
dtd
IsraTrance Junior Member
Started Topics :
17
Posts :
490
Posted : Jun 18, 2006 15:29
|
|
trip: typically in such applications, the auditory features extracted from audio data are used to train statistical pattern recognition classifiers. by training them with a large set of no.1 hits from the past, the classifier can build a concept model describing a no.1 hit based on the extracted features. then it can decide on novel patterns if they are in its concept or not. this is scientifically interesting, i think. whether or not it is used to determine no.1 hits, it serves as a training example to learn more about how audio is percieved, classified and 'rated' by humans. of course, the article contains some amount of 'economist view' and is somewhat imprecise.. |
|
|
|
dtd
IsraTrance Junior Member
Started Topics :
17
Posts :
490
Posted : Jun 18, 2006 17:18
|
|
an interesting application is automatic music retrieval. by training a classifier with samples from your favourite music collection (here, samples are not the samples from a/d conversion but samples in statistics), the classifier could be able to serve you with novel music currently released, hopefully matching your likings. this is currently researched, do you like this idea? it's very interesting to me.. |
|
|
|
Trip-
IsraTrance Team
Started Topics :
101
Posts :
3239
Posted : Jun 18, 2006 17:53
|
such as Pandora or the likes...?
again, as insteresting as it may be, classifying number 1 hits is probably useless, since most of it is simply music for the masses... And that is mostly a money maker than a music maker.
Crackling universes dive into their own neverending crackle...
AgalactiA |
|
|
|
dtd
IsraTrance Junior Member
Started Topics :
17
Posts :
490
Posted : Jun 18, 2006 18:59
|
|
yes, if it works, it will be probably used for commercial purposes. but it is important to note that the classification results depend on the training phase (and the extracted features), as classifiers are general enough to be used in different contexts. therefore, they can be as well used only on specific music genres, not only on 'music for the masses'. imagine to train it using only psytrance music. the classifier will then build a concept model of 'no.1 psytrance' music (whatever is meant by 'no.1 psytrance' of course...) |
|
|
|
