13 December 2004

Feed integrators: How do you do it?

I was talking to someone about a "feed integrator" [another one of those terms I like to use, and maybe I'm the only one using it.].

Feed integrators take multiple feeds and mix them. It's not an aggregator. It's more interactive.

Think of feeds as DNA strands. And the individual blog entries or new articles are things that "get mixed up" in the integrator.

At the "output side" of the integrator is a new feed, re-prioritized, and new patterns.

The idea of an integrator isn't to simply showcase feeds in new orders. It's to find new relationships among the individual feed-elements, and then spot new patterns and trends.

For example, suppose you are interested in traffic. You might take in your traffic reports as one feed; another feed might be the history of accidents along your rate; and a third feed might be the simulation outputs form your linear programming model on particular routes; a fourth feed might be the comparison between the simulated vs actual backlogs times; and a fifth feed might be anecdotes and traveler reports along your path.

You've got all this stuff flying in. The idea of an integrator is to see the new patterns across the five feeds and then guide you to "what is the most appropriate route."

Sure, one solution might simply to rename the "feed integrator" something like an operating system. But why not let the feeds interact on their own outside your computer.

Feed integration could also be applied to personal choices by combining various feeds about risks in the environment; indicators about certain economic trends; and ongoing preparation plans to support these activities. The integrator would evaluate the situation, and layout the decisions that need to be made.

The question becomes, "How do we do this integration?"

Well, look at statistics. You can take one item from a bag, and have certain finite combinations. The trick with a computer is that you can pick, compare, and recombine the combinations quickly.

The steps build down to picking the blog-entry, finding the pattern, differentiation, and identifying a logical flow.

Sure, it's not easy. But that's why people get paid to make things that don't already exist.
I was talking to someone about a "feed integrator" [another one of those terms I like to use, and maybe I'm the only one using it.].

Feed integrators take multiple feeds and mix them. It's not an aggregator. It's more interactive.

Think of feeds as DNA strands. And the individual blog entries or new articles are things that "get mixed up" in the integrator.

At the "output side" of the integrator is a new feed, re-prioritized, and new patterns.

The idea of an integrator isn't to simply showcase feeds in new orders. It's to find new relationships among the individual feed-elements, and then spot new patterns and trends.

For example, suppose you are interested in traffic. You might take in your traffic reports as one feed; another feed might be the history of accidents along your rate; and a third feed might be the simulation outputs form your linear programming model on particular routes; a fourth feed might be the comparison between the simulated vs actual backlogs times; and a fifth feed might be anecdotes and traveler reports along your path.

You've got all this stuff flying in. The idea of an integrator is to see the new patterns across the five feeds and then guide you to "what is the most appropriate route."

Sure, one solution might simply to rename the "feed integrator" something like an operating system. But why not let the feeds interact on their own outside your computer.

Feed integration could also be applied to personal choices by combining various feeds about risks in the environment; indicators about certain economic trends; and ongoing preparation plans to support these activities. The integrator would evaluate the situation, and layout the decisions that need to be made.

The question becomes, "How do we do this integration?"

Well, look at statistics. You can take one item from a bag, and have certain finite combinations. The trick with a computer is that you can pick, compare, and recombine the combinations quickly.

The steps build down to picking the blog-entry, finding the pattern, differentiation, and identifying a logical flow.

Sure, it's not easy. But that's why people get paid to make things that don't already exist.
" />