Likes and Loves

By | July 21, 2009

HypeM is attempting something pretty cool. They wondered if it was possible to create a music chart driven by people for people, one that actually represents a diverse set of music and interests.

It’s not that HypeM doesn’t already do a pretty good job of this with their “popular” list of songs marked with big red hearts, those “songs scoring the most favorites on The Hype Machine,” by users like you and me.

But with Twitter, they thought, you could potentially have a music chart generated by anyone who links to a HypeM song, no need to have an account with HypeM.

This took a couple of brilliant decisions on their part. First, they needed to value each Twitter user, which they’ve taken a good shot at with this formula (. . . and already trying to better it by asking for feedback from the math geeks):

round(( 1/3 * (twitter_followers / 10) ^ 0.5 ) * (twitter_followers / twitter_friends) * 10))

The formula weighs individual influence and group influence and puts a ceiling on how much influence one or the other can have. But, the key point is that I have a Twitter score, you have a score, anyone can have a score.

And when you tweet a song that is on HypeM, you and your score are added to a group for that particular song on the Twitter Music Chart. It rises based on the sum of those scores.

Very cool, but HypeM also took another step to make this chart even better, deeper. They used BackType, which aggregates comments from all over the Web, to suck in and interpret all of the tweets out there before using them. The result is that when someone uses a URL shortener (like to link to a HypeM song, BackType recognizes it and includes it in the useful pool of tweets.

Why care about this way of building a music chart? Because what people like, love, and discuss on the open Web, all of those needles in the haystack, needs to be gathered up to create something that didn’t exist when those sentiments existed by themselves. HypeM just gave us a great example of how to do it.



  • Taylor Davidson

    This is brilliant; honestly I hadn't recognized the implications or potential application until you pointed it out. It's algorithmic scoring of the value of a bit of content, using social input, in real-time.

    But that's an unnecessarily geeky way to say it.

    Now, imagine:
    - a similar system for any type of content
    - if the chart was customized for me based on the same inputs but their relation to me; possible to compute, but incredibly hard to execute at scale

    • brooksjordan

      Yep an “algorithmic scoring of the value of a bit of content, using social input, in real-time,” nicely said.

      Now, how could the chart be customized so that the inputs related to me or you?

      • Taylor Davidson

        One of many attempts to measure value using authority and popularity; at it's heart, that what PageRank does, just with different inputs, on a different schedule, with a different output, right?

        How could the chart be customized?

        I may care about authority and popularity different than other people; in fact, I may only care about a song's popularity within my social circle, or I may only care about certain authority figures: particular critics, friends, or people with notably good / similar taste.

        I may want my chart to reflect what you care about and rank your vote higher than someone else's vote, simply because I like your taste, despite the overall rankings.

        It's just a more algorithmic approach to the social filter we have always and will always use.

        Is it computationally possible? Yes, but hard to do at scale. Is it truly valuable? Dunno.

        • brooksjordan

          Mm, very intriguing, “page rank” for social content.

  • viicOop

    wOw:!..U ar3 a geek:!..Lolzz
    ('am saying iit iin a gOod waii ;)

    • brooksjordan

      Are you one, too, viicOop? You're practically writing in Perl, right? :)

      • viicOop