rndm thghts

As I’m an iPhone developer I follow a lot of mac heads, so I’ve been suffering the iPad launch. This got me thinking of ways to avoid it without missing out totally.

I’m lazy so I did not make any research about this. I’m sure this might have already been studied in depth, anyway, here it goes.

The objective could be to let the truly original information get through, collapse the duplicates within the same tweet (so you can see it anyway if you want), and at the same time generate some kind of score about how many people tweeted that particular message to know how important it is. The meaning of important is implied in who you follow (except the difference between personal and the rest, that is not clearly defined in tweeter, but that’s to discuss in other time), so the numbers will be different for each twitter user and is not applicable to the firehose.

A way the filter might work is to let the first tweet that contains certain information shows up. The information score I=1, and the popularity P=1 too. A second tweet comes with the “same” information, so now P=2, but the I is less, let’s say I=1/2. A possible simplification is I=1/P. This second tweet is also linked to the first one, so it can be collapsed under it by the client.

I could ask the filter to:

• show me only the truly original info: shows those with P=1
• show me those that were at least important for 2 of my friends: shows tweets linked to those with P=2

This filter can be made in realtime by the clients, wheere you define the time window (e.g. 24hs). This limits the resources you need too.

One possible use of this as a custom filter is that the “first tweet” is actually typed in the client by the user. Today, for example, I could have added “iPad arrived”, “iPad unboxing”, “playing with iPad”, “love iPad” and skipped the envy. :)