Published on 18 November 2012
An interesting post ran across my twitter feed this morning:
cortesi: I want a tool that calculates the cost of following a Twitter user. Something like volume* signal/noise…
I started thinking that this is something you could do automatically. Cortesi seems to already be working on a side project that would add a like button to tweets, so that he can manually manage which ones he likes and help the program predict more. I think that this is a good way to go, but the data deluge may get exhausting training wise, and it depends upon too much of a conscious choice on the part of the tweeter.
Instead, it might be worthwhile making a program that logs:
Combined, this should work fairly well to let the app know what sort of information you like on your feed. It doesn’t work for self-contained tweets which you enjoy, however - in that case, the manual system, worked out by cortesi, would be best. But still, this would be a major bootstrapper.
I’m thinking about how I would do it. I would log links followed using a browser add-on - I’m not sure if Twitter development code allows you to log followed links (anyone know?) I would also do a simple reply search to log that, again using the Twitter api. And then perhaps some basic NLP on those posts to see what sort of thing I’m replying.
How would you do it?
Tagged: