2006

25/6

Uncategorized

Music You Like

by | Print
{mosimage}

Ashwin Raghu has been playing around with Pandora, the revolutionary new internet radio player. 

You start off wanting to listen to some Bob Dylan. You are served up a couple of anecdotes and that screeching harmonica. There is a pleasant smattering of folk that follows: Woody Guthrie and Hank Williams and Arlo Guthrie. Dylan then comes back on with a bluesy number. You like the variation, and what follows is a stretch of music that is still the kind you started off wanting to listen to, only now with a touch of the blues. Along the way you discover artists you may have never heard of, like Tom Waits, whose low-down gravelly voice rivals that of Lou Reed. And Townes Van Zandt, who, his profile tells us, had a “hard enough time simply keeping his records in print", something that seems inexplicable. And then choices you’re surprised by, like Belle & Sebastian, one of those highly acclaimed ‘new’ bands that you’ve never gotten around to listening to.

 

It is this musical serendipity that people who use Pandora are charmed by. Pandora.com wants to be your personal DJ, and asks you to begin by choosing a song or artist that you like. Every subsequent song that is picked to be played on that station will be ‘related’ to your original choice in some way. The Music Genome Project that drives Pandora deconstructs each song and assigns it a set of ‘genes’, based on its musical qualities. The feature that gives them authenticity, indeed the cornerstone of their "We want to have a conversation with you" theme is that the genes for each song are a result of a bunch of Pandora’s in-house musicians listening to and classifying each of the thousands of songs that are now on the database. For instance, ‘Dirrty’ by Christina Aguilera is described as ‘vocal-centric’, having ‘minor-key tonality’, ‘high synth use’ and ‘repetitive melodic phrasing’. And this is only scratching the surface, the most significant of the four hundred genes for that song.

 

What is interesting is that the project seeks to classify the music completely independent of who the band is, how popular they are, who they were influenced by, the period that their music belongs to or where they are from: all hooks that we are used to going by when we choose our playlists and CDs. The fact that the particular song you are listening to is, say, a Reggae number out of the UK in the mid-1970s has no bearing on its genes. The focus is purely on its inherent musical attributes. Create a station with Britney Spears’ ‘Toxic’ and you will soon be listening to other songs with ‘House roots’; songs that you might have never associated with Britney otherwise. A hip-hop station will not bother with whether the rapper was schooled on the East Coast or on the West Coast, nor will it give any credence, refreshingly, to the rapper’s allegiance as regards the Biggie / Tupac rivalry.

 

And you can experiment! Those musical fantasies of ‘A little less Pearl Jammy would be nice, and it would be perfect if the vocalist had some of R.E.M.’s out there-ness’ type stuff might be closer to reality, and fruition, than you think. And if you fiddle around long enough, you just might find the perfect mix, considering Pandora seeks out indie and local bands to send in their music as much they do the Billboard stuff. You can tweak a station by suggesting other artists that you think should be part of the mix. Pandora responds by suggesting music that it thinks embodies qualities of both the artists you’ve chosen. Each session of listening to Pandora can thus turn into quite a trip; one that you have just the right amount of control over.

 

There are issues though. The most glaring of which is when songs end up repeating within short intervals. This is very surprising, considering the size of their database (half a million tracks and counting) and the multi-dimensionality of the criteria that go into choosing a song. Some genres of music are better represented than others. There is no classical music, for instance: Ludwig van would be quite amused to know that a cursory search for "Beethoven” leads to a song called "Ain’t Nothin’ But The Blues". Their collection of all-encompassing World Music is still fledgling, a collection that will no doubt grow.

 

Pandora aspires to reach out to you on a personal level; their warm-fireplace user-friendliness really encourages feedback. You want to get involved in creating your perfect radio station. But is it as personal as writing in to the Radio Jockey at your favourite “manned” internet radio station? (Let’s not talk about terrestrial radio programming in India, for many reasons; a stultifyingly tiny catalogue is the first thing that comes to mind.) My favourite until I became captivated by the music-DNA-magic of Pandora was live365.com, a more ‘traditional’ online federation of radio stations with a human RJ playing music that he liked, and wanted other people to listen to. Admittedly it takes some sifting through to find the wheat among all the chaff, but once you find a Radio Jockey who ‘gets’ your music preferences, he becomes somebody with whom you can talk to about what kind of music you would like to hear and what kind of music you think would ‘fit’ the station.

 

Which brings us to what is and what will constantly be Pandora’s biggest challenge. To what extent, and in what way, do musical ‘genes’ actually correlate to how the average listener perceives a piece of music? A function that is even more important to a radio station than playing the right song is playing the right set of songs. Can genes determine if one song will segue well into another? If so, will they be able to contribute to and enhance the overall flow that is expected of a station? Can a set of musical attributes, even up to four hundred of them, translate to deciphering our ‘mood’ when we listen to a song? But maybe it is unreasonable to even expect this. What Pandora seeks to achieve is to arrive at a set of musical guidelines that can come as close to this as possible and appropriate it to the millions of users that they hope to reach. This is something that they have managed to do brilliantly.

(Ashwin Raghu likes to think about music when he’s not listening to it. His Fab Four would be Robert, Jimmy,John Paul and John, although Roger, David, Nick and Richard would run them close. Just as John, Paul,George and Ringo would. Looking back at that, he’s pondering the possibilities of a supergroup with John,John, John and John, and not necessarily in that order
either. Did he say "think"?!)

{mosimage}

({mhauthor})

Also by

Share This

2006

25/6

Uncategorized

Music You Like

by | Print
{mosimage}

Ashwin Raghu has been playing around with Pandora, the revolutionary new internet radio player. 

You start off wanting to listen to some Bob Dylan. You are served up a couple of anecdotes and that screeching harmonica. There is a pleasant smattering of folk that follows: Woody Guthrie and Hank Williams and Arlo Guthrie. Dylan then comes back on with a bluesy number. You like the variation, and what follows is a stretch of music that is still the kind you started off wanting to listen to, only now with a touch of the blues. Along the way you discover artists you may have never heard of, like Tom Waits, whose low-down gravelly voice rivals that of Lou Reed. And Townes Van Zandt, who, his profile tells us, had a “hard enough time simply keeping his records in print", something that seems inexplicable. And then choices you’re surprised by, like Belle & Sebastian, one of those highly acclaimed ‘new’ bands that you’ve never gotten around to listening to.

 

It is this musical serendipity that people who use Pandora are charmed by. Pandora.com wants to be your personal DJ, and asks you to begin by choosing a song or artist that you like. Every subsequent song that is picked to be played on that station will be ‘related’ to your original choice in some way. The Music Genome Project that drives Pandora deconstructs each song and assigns it a set of ‘genes’, based on its musical qualities. The feature that gives them authenticity, indeed the cornerstone of their "We want to have a conversation with you" theme is that the genes for each song are a result of a bunch of Pandora’s in-house musicians listening to and classifying each of the thousands of songs that are now on the database. For instance, ‘Dirrty’ by Christina Aguilera is described as ‘vocal-centric’, having ‘minor-key tonality’, ‘high synth use’ and ‘repetitive melodic phrasing’. And this is only scratching the surface, the most significant of the four hundred genes for that song.

 

What is interesting is that the project seeks to classify the music completely independent of who the band is, how popular they are, who they were influenced by, the period that their music belongs to or where they are from: all hooks that we are used to going by when we choose our playlists and CDs. The fact that the particular song you are listening to is, say, a Reggae number out of the UK in the mid-1970s has no bearing on its genes. The focus is purely on its inherent musical attributes. Create a station with Britney Spears’ ‘Toxic’ and you will soon be listening to other songs with ‘House roots’; songs that you might have never associated with Britney otherwise. A hip-hop station will not bother with whether the rapper was schooled on the East Coast or on the West Coast, nor will it give any credence, refreshingly, to the rapper’s allegiance as regards the Biggie / Tupac rivalry.

 

And you can experiment! Those musical fantasies of ‘A little less Pearl Jammy would be nice, and it would be perfect if the vocalist had some of R.E.M.’s out there-ness’ type stuff might be closer to reality, and fruition, than you think. And if you fiddle around long enough, you just might find the perfect mix, considering Pandora seeks out indie and local bands to send in their music as much they do the Billboard stuff. You can tweak a station by suggesting other artists that you think should be part of the mix. Pandora responds by suggesting music that it thinks embodies qualities of both the artists you’ve chosen. Each session of listening to Pandora can thus turn into quite a trip; one that you have just the right amount of control over.

 

There are issues though. The most glaring of which is when songs end up repeating within short intervals. This is very surprising, considering the size of their database (half a million tracks and counting) and the multi-dimensionality of the criteria that go into choosing a song. Some genres of music are better represented than others. There is no classical music, for instance: Ludwig van would be quite amused to know that a cursory search for "Beethoven” leads to a song called "Ain’t Nothin’ But The Blues". Their collection of all-encompassing World Music is still fledgling, a collection that will no doubt grow.

 

Pandora aspires to reach out to you on a personal level; their warm-fireplace user-friendliness really encourages feedback. You want to get involved in creating your perfect radio station. But is it as personal as writing in to the Radio Jockey at your favourite “manned” internet radio station? (Let’s not talk about terrestrial radio programming in India, for many reasons; a stultifyingly tiny catalogue is the first thing that comes to mind.) My favourite until I became captivated by the music-DNA-magic of Pandora was live365.com, a more ‘traditional’ online federation of radio stations with a human RJ playing music that he liked, and wanted other people to listen to. Admittedly it takes some sifting through to find the wheat among all the chaff, but once you find a Radio Jockey who ‘gets’ your music preferences, he becomes somebody with whom you can talk to about what kind of music you would like to hear and what kind of music you think would ‘fit’ the station.

 

Which brings us to what is and what will constantly be Pandora’s biggest challenge. To what extent, and in what way, do musical ‘genes’ actually correlate to how the average listener perceives a piece of music? A function that is even more important to a radio station than playing the right song is playing the right set of songs. Can genes determine if one song will segue well into another? If so, will they be able to contribute to and enhance the overall flow that is expected of a station? Can a set of musical attributes, even up to four hundred of them, translate to deciphering our ‘mood’ when we listen to a song? But maybe it is unreasonable to even expect this. What Pandora seeks to achieve is to arrive at a set of musical guidelines that can come as close to this as possible and appropriate it to the millions of users that they hope to reach. This is something that they have managed to do brilliantly.

(Ashwin Raghu likes to think about music when he’s not listening to it. His Fab Four would be Robert, Jimmy,John Paul and John, although Roger, David, Nick and Richard would run them close. Just as John, Paul,George and Ringo would. Looking back at that, he’s pondering the possibilities of a supergroup with John,John, John and John, and not necessarily in that order
either. Did he say "think"?!)

{mosimage}

({mhauthor})

Also by

Share This

Comments

Leave a Reply




Close
E-mail It