Songwriters & Musicians & Demographics

[kfbkfb]

Maybe a 3rd party (AI?) based computer system that listens to a radio station for weeks and weeks.

The AI listening system might detect a commonality in the songs (crashing cymbals, screaming synthesizers, booming bass, very emotional [sounding] vocals) over the weeks and weeks of monitoring, this could be matched to the ratings (and demo breakdowns) over those same weeks and weeks and be sold to composers as a guide for what sort of song content is likely be popular on that radio station.

Although this method would limit creativity, it would increase the likelihood of the composer(s)/singer(s)/musician(s) making a profit from their work.


Kirk Bayne

 

[davideduardo]

Maybe a 3rd party (AI?) based computer system that listens to a radio station for weeks and weeks.

The AI listening system might detect a commonality in the songs (crashing cymbals, screaming synthesizers, booming bass, very emotional [sounding] vocals) over the weeks and weeks of monitoring, this could be matched to the ratings (and demo breakdowns) over those same weeks and weeks and be sold to composers as a guide for what sort of song content is likely be popular on that radio station.

The problem is that ratings are only moderately granular, with 15 minute chunks being the shortest measured period. In 15 minutes of a music sweep, you can have four or five songs.

And Nielsen data only shows detections. If a person's meter does not detect for a few minutes, there can be many reasons, only some of which might be tune out due to a disliked song; more likely it is due to going to the bathroom, arriving at work, getting a phone call, etc.

So it would take an entirely different system that could correlate minute by minute listening to music from radio, streams and such with individual behavour. And that still would not measure accurately due to so many other reasons than the music itself for a person to stop listening to a song.

 

[TheBigA]

Maybe a 3rd party (AI?) based computer system that listens to a radio station for weeks and weeks.

That's what Mediabase and Billboard do. They monitor all reporting radio stations and print a chart each week that lists the most played songs into a Top 50. The charts are weighted based on radio station audience and ratings. So there you go.

But they don't get critical about the music itself because it's an unemotional, uninvolved computer system. All they do is listen for the codes that identify the songs and count them. You're asking a computer to understand music and human emotion. It does not compute. It's up to humans to look at that information, listen to the most popular songs, and determine if there's a pattern or commonality among them. I suspect that there is.

 

[davideduardo]

That's what Mediabase and Billboard do. They monitor all reporting radio stations and print a chart each week that lists the most played songs into a Top 50. The charts are weighted based on radio station audience and ratings. So there you go.

But they don't get critical about the music itself because it's an unemotional, uninvolved computer system. All they do is listen for the codes that identify the songs and count them. You're asking a computer to understand music and human emotion. It does not compute. It's up to humans to look at that information, listen to the most popular songs, and determine if there's a pattern or commonality among them. I suspect that there is.

Songs are not encoded. It's actually the opposite: BDS and Mediabase create a unique computer signature for each song using a complex algorithm which codifies the sound of the song in a way that is unique. So I wonder if common elements of hit songs could be analyzed with a resultant conclusion or formula. I suspect not.

I suppose that an algorithm could be developed to determine if there is any uniformity in when a listener tuned out of a song and then analyzed a variety of factors right down to individual notes and bars. But I'm not sure that this would be any more successful than trying to learn how to paint like Michelangelo by analyzing his brush strokes.

 

[TheBigA]

Songs are not encoded. It's actually the opposite: BDS and Mediabase create a unique computer signature for each song using a complex algorithm which codifies the sound of the song in a way that is unique.

Here's what BDS says. "Nielsen Broadcast Data Systems tracks radio airplay using a patented digital
pattern recognition technology." That sounds like encoding to me.

In fact when you submit music to BDS, you send it to "virtual encode."

Regardless, the only thing they do is count the number of spins songs get.

 

[davideduardo]

Here's what BDS says. "Nielsen Broadcast Data Systems tracks radio airplay using a patented digital
pattern recognition technology." That sounds like encoding to me.

No, it is not encoding. They take each song and the algorithm analyzes it and essentially finds its "fingerprint" based on all kinds of factors like key, note sequences, frequency of sequences, speed of sequences, tones of voice and instruments and the combination of them and so on.

The simplest and best analogy, to repeat the term, is a fingerprint. No two are identical.

In fact when you submit music to BDS, you send it to "virtual encode."

Yep, that means a clean copy, that is all. Then the computers build its unique identification.

As a song is "heard" by BDS or MediaBase monitors, every microsecond reduces the possible songs that are similar until the actual one is identified. While the services require over a minute of song play for a credit to avoid commercials or buffer music getting recognized, it only takes a few seconds of a song to find it's uniqueness.

Regardless, the only thing they do is count the number of spins songs get.

Absolutely. All they want is artist and title and station and time.