Tweeting with #Heston – Mapping the Celebrity Chef Twitterati
Watching Heston Blumenthal doing his stuff with eggs* earlier this evening (note to self: buy dry ice), and idly pondering who the equivalent personality might be in I(C)T, given Govey’s ramblings today, I wondered how useful a probe using the #heston tag might be into the world of celebrity cooking…
(* crap URL chaps: http://www.channel4.com/programmes/how-to-cook-like-heston/episode-guide/series-1/episode-1 for Episode 2; or are we getting folk into the idea that an index count starts at 0?!)
…so I did a search for recent users of the #heston hashtag, filtered the list to only include folk who had sent two or more correspondingly tagged tweets in my sample, and then graphed them along with connections to their friends. Filtering this graph to include nodes with in-degree of 20 or more (so at least 20 followers amongst the filtered hashtag users) (inclusive-)or out-degree of 10 more (so each sampled tag user had at least 10 friends) produced this result:
By removing the filter requirement on using the tag at least twice in the sample of tweets I captured, we get an audience with potentially wider set of interests. Generating the friends network and then setting quite a high generic filter threshold (save nodes with degree 75 and up), we map a broader range of interests (where people followed are a proxy for interest):
Using the same filter parameters as in the first example (indegree: 20, outdegree: 10, though it must be remembered the orginal hashtag user base sample size is now larger), this is what we get:
Top right are the chefs, bottom right are the pop intelligentsia, bottom left are the generic celebrities.
These maps are still pretty crude, but they do demonstrate how easy it is to fish out a good selection of celebrity cooking related twitter accounts from a loosely related hashtag. Here’s a close-up of the cooking related cluster:
It’s easy enough to profile areas using a variety of seeds: hashtags, as in this example, specific twitter users (profiled either according to who they follow, or who their followers follow), sets of users (or Twitter lists), search terms, URLs, locations (actually, I haven’t implemented the location based search in my current script)…
The thing is, it feels interesting, but is it useful?!