Title *

SummaryPart of this ongoing research is reflective, I'm trying to understand better the idea of connecting together people over a geographical landscape.

Description *To help keep the project focused I'm going to try to visualize the signaling space over Portland. The hope is to aggregate data from twitter, to cluster related articles, and to visualize the the active connections or potential connections as they happen in real time. Producing a visualization is a three step process: 1) Aggregation. For test data I am now aggregating all twitter traffic over Portland Oregon. This is just using a cron job with the ruby twitter gem - and it was a piece of cake. 2) Clustering. To cluster related posts I first tried writing my own traditional search engine - this works well and is currently mounted at http://citybot.org/search . But this doesn't really address the issue of clustering related posts. So instead now I am trying out the carrot2 search engine. As well I am considering throwing out my aggregation engine and replacing it with nutch - which would scale better. 3) Visualization. As far as visualization itself I am trying processing - where I've written a small engine that can visualize tweets with some special effects over time - onto a map. This wasn't terribly hard; the real challenge it turns out is the middle phase - of clustering related posts. As soon as I have some good data to render I should be able to do so. There is some small argument of switching away from processing over to papervision3d - but it is a hassle to build a compiling environment for flex and that still requires research so I am sticking with processing for now. This new work and time I've put in overall has given me some new perspective lately. As well changes in the market are also changing my perspective. Several twitter based market place applications have emerged - and we do see the rise of twitter agents in a way similar to IRC. On the web two new services have emerged: http://hutch.com and http://vark.com - both of which are effectively a kind of social search. I very much want a "real time" sense to citybot. I'd like to let you ask a question of your geography essentially. The reality of social engagement is not one of matching complementary needs against each other but rather matching needs against people in your community. My overall feelings now are that: 1) The real value of a social search service is to help you find the person who can help you - not just find the data but find a person - find a party to have a dialogue with. Perhaps this is obvious but it bears keeping in mind. 2) I still feel that surfacing and sharing those implicit publishing moments ; the kinds of moments that google collects whenever you do a search - is vital. This is something I really learned from Meedan. 3) It's ok to connect communities of people who are all seeking the same answer - because it helps to crowd-source the problem of finding those answers. I had previously been thinking that one wanted to only connect complementary interests. 4) Do you have to message some agent when you have a need? This is an ongoing but evolving question. Perhaps the biggest shift in my thinking has been that one should not bother requiring an agent that you talk to by messaging it directly. Rather I think I should just measure the happiness or satisfaction over Portland ( by looking for satisfaction terms in twitter such as 'thanks!' ) and then apply my agent to that traffic, to help create satisfaction moments, and then to measure again to see if I've improved the number of satisfaction events. In this way I'm tackling empirically measurable issues ( quality of life ) and testing for real results. I'm looking to have a quiet quality of life improvement; one that doesn't necessarily translate to increased GDP or Standard Of Living or other such economic measurables - and in fact probably works against them. 5) In all this work my emphasis lately is to try and facilitate human brokers. The hope on the website is to let people match-make other people and to get credit for that. This means that the clustering tools and the like have to appear in a place where an expert can use them to browse the posts that people are making and find likely matches.