On Thu, 18 Mar 2010, Richard Pope wrote:
Can anyone think of a way of spotting the seeds of those agendas? I'm guessing some kind of measure of connectivness + repeated phases might do it. Would need to be near realtime and have a good alarm system when / if it spotted stuff. Is that even barely realistic of would it require too much magic?
post-process it? You can probably see what gets popular, and then step backwards through time; and look at where the same thing came from multiple sources in a similar way without obvious connections based on follows/interests and also what else was/wasn't tweeted between suspected clusters. It's not quite near real time to develop the suspect accounts, but if they'll will be reused through the programme, you can do something if they tweet something similar at the same time, which doesn't match what the majority of discourse is saying. I can see that also working for pre-planned attack lines, where one party pushes a series of tweets out to see if it can be got trending, knowing it'll be used later. It may need a twitter firehose feed to do properly; not sure. Sure would be a fair bit of API traffic. Cheers Sam -- We are not here to keep people from being stupid. Once they are stupid, we will pick up the pieces. _______________________________________________ Mailing list [email protected] Archive, settings, or unsubscribe: https://secure.mysociety.org/admin/lists/mailman/listinfo/developers-public
