Ah, then it sounds like you need The Firehose (tm). Or ask the twitscoop guys how they're getting their data (which I am now curious about). -chad
On Mon, Feb 9, 2009 at 1:50 PM, Jennie Lees <trin...@gmail.com> wrote: > > Aye, that'd work if I knew the terms in advance, but I'm hoping to do > something more dynamic that could work for any term. Thanks for the > suggestion though! > > Jen > > On Mon, Feb 9, 2009 at 4:23 PM, Chad Etzel <jazzyc...@gmail.com> wrote: >> >> If you are just looking for keeping track of the number of tweets, and >> not concerned about keeping them around for later data processing, you >> can poll the search API every so often for your terms of interest and >> just update your counters on the fly. >> >> pseudo code: >> >> let TOTAL = 0 >> let BUCKET = array of timeslices >> >> Do Forever >> Poll Search API for <term> >> let C = number of results >> TOTAL := TOTAL + C >> let BUCKET[i] be the current timeslice bucket (e.g. 11-12) >> BUCKET[i] := BUCKET[i] + C >> Store TOTAL and BUCKET[i] back to database >> Sleep for a bit >> End Do >> >> >> Not sure if this would be close to fitting your needs, but something >> to start with maybe... >> -Chad >> >> On Mon, Feb 9, 2009 at 10:52 AM, Jennie Lees <trin...@gmail.com> wrote: >>> >>> Hi all, >>> >>> I'm looking for a way to find the number of tweets matching a certain >>> term, both overall and as a time series (e.g. how many results between >>> 11:00 and 12:00 - I want to monitor changes over time). >>> >>> I can't really see an obvious way to do this beyond making a local >>> copy of every tweet and searching those, then counting the results. I >>> believe that's what TwitScoop does, but I can't help but think this >>> isn't the most efficient way. >>> >>> Has anyone found another approach? >>> >>> Jen >>> >> >
