What insight is there in geolocated tweets? 1

Thanks to agoasi https://www.flickr.com/photos/bartholl/

In his recent post on geotags in twitter Thierry Gregorius said

“If you use Twitter then you will have noticed that the service encourages people to geocode their tweets, that is, to record their physical location at the moment of tweeting. What particular purpose this may serve is another point altogether, but let’s not get into that.”

Well I do want to “get into that”

With so called smartphones it is easy to add an x,y to almost everything, photos, tweets, checkins, blogposts. For many of us it is easier to leave geotagging on rather than dive into the deep settings to switch on/off selectively – Warning extensive use of GPS may run down your battery.

I have been wondering what is the point of geotagging tweets? Some of you may have seen some early thoughts on this from last year’s W3G Conference/Unconference and nearly a year on I have yet to find any convincing uses of geotagged tweets.

Several academics have looked into the topic and to the best of my knowledge no one has come up with any meaningful relationships between the content of a tweet and the location from which it is tweeted. It seems to me that the best that we can get from geotagged tweets is the locations and times of activity of people with mobile devices who use twitter, probably corresponds to a digi/socialmedia/techy demographic and largely urban.

This demo from geo.me let’s you search for any hashtag or term on twitter (select tweet mapping) and this one from the last election shows political boundaries so you can see if you can find any political trends in the tweets that your search returns. Maybe someone will spot a significant correlation.

My friends at GeoIQ have done some pretty awesome stuff using a sentiment analysis engine on the twitter feed from this year’s Oscars to visualise the content, see Sean Gorman’s blog post and the visualisation. Sean acknowledges some of the limitations of the analysis

A second challenge with location based sentiment analysis is how meaningful are the results. I think one of the things we miss are margin of error calculations for sentiment analysis. Once we’ve aggregated data we have a sample size for that geography that we can calculate a margine of error against.

This is the best that I have seen so far, but does it really provide much in the way of insight? I am not sure.

Anyway for the moment the “locate my tweets” feature is switched off, it’s enough that I bombard you with my opinions without sharing where they originated from. Just in case you had forgotten I also still have some reservations about the privacy of broadcasting my location, show me my gain and I might change that view though.

One thought on “What insight is there in geolocated tweets?

  • steven

    Ken Field and James O’Brien have pointed me at their paper “Exploring cartographic design in social network map mashups” at https://bit.ly/omFFdE

    As the title states the paper is primarily focussed on techniques to represent the data points represented by geotagged tweets rather than considering what insight can be derived from a spatial analysis of the data. This is particularly true in my opinion when the twitter content is being passively sourced (i.e the tweeter is not consciously structuring the content of the tweet to convey information specifically related to the location from which they are tweeting). However the field work case study at the end of the paper illustrates twitter as a useful tool when used in a “purposeful” scenario.

    I particularly enjoyed the quote from Carl Steinitz “much of it is visualisation but not
    communication…there is little point to much of it and it’s not particularly ‘useful’. It’s made worse because there is a preoccupation with making things go round and round and up and down…worse if it’s
    accompanied by music” I recall sitting next to Ken at that lecture and we were both roaring with laughter and frenziedly tweeting it out (can’t remember if the GPS was on)

Comments are closed.