SoMe Lab Social Medial UW Geolocation and Geographic Imaginaries (or place framing as the “becoming” of networked space) Josef Eckert University of Washington Feb. 24, 2012
SoMe Lab Social Medial UW Making sense of this…. from the flickr stream of _PaulS_
SoMe Lab Social Medial UW As it relates to this. Occupy-related corpus, hex-binned (2dd), geolocated tweets, Oct. 19 – Dec. 31
SoMe Lab Social Medial UW Full disclosure
SoMe Lab Social Medial UW The fuzziness of physical, relational, and networked space Protest as contributing to the “becoming” of space, negotiating “the right to the city” (Lefevbre, Mitchell 2003) “Place-framing as place-making” (Martin 2003) The expanding role of technology and networks for social movements and resistance within space (Nicholls 2008 a,b ) Tobler’s Law (1970): “Everything is related to everything else, but near things are more related than far things”
SoMe Lab Social Medial UW Collection Method (AWS & Hadoop)
SoMe Lab Social Medial UW The Baleen Whale Approach to Data Gathering Twitter is ephemeral 200 million tweets per day 178 related search terms via Twitter API ~300,000 – 1,000,000 tweets per day collected “The edges of these bones are fringed with hairy fibres, through which the Right Whale strains water, and in those intricacies he retains the small fish, when open-mouthed he goes through the seas of brit in feeding time.” (Melville 1851)
SoMe Lab Social Medial UW Seeding keywords: occupytogether.org Number of Occupiers Reporting on Meetup.com Oct. 26
SoMe Lab Social Medial UW Occupy-related corpus, hex-binned (.05 dd), geolocated tweets, Oct. 19 – Dec. 31
SoMe Lab Social Medial UW New YorkOaklandSeattle 1,480 (0.03%) Clustered, p< (37.09%) 791 (0.02%) Clustered, p< (46.65%) 199 (0.004%) Clustered, p< (47.70%) Within city ANN Within 1000 ’
SoMe Lab Social Medial UW But that’s not a lot of tweets! Is there any other way we can read geographies from Twitter activity? Is 20,645,921 tweets enough of a start?
SoMe Lab Social Medial “Elderly woman takes pepper spray to the face #ows #occupyseattle #occupypdx”
SoMe Lab Social Medial UW Measuring great circle distance
SoMe Lab Social Medial UW Hashtag co-occurrence great circles (unclassed), Oct. 19 – Dec. 31 ~6,700 links
SoMe Lab Social Medial UW Bias of the geo-crowd (Zook & Graham 2010) = Biased data sampling
SoMe Lab Social Medial UW A raw count of co-occurrence
SoMe Lab Social Medial UW Not significant Linear regression to determine explanatory power of distance for hashtag co-occurrence. Not even close
SoMe Lab Social Medial UW Jaccard Index A ∩ B A ∪ B
SoMe Lab Social Medial UW Which looks like this…. Austin, TX Houston, TX Memphis, TN Milwaukee, WI Auburn, CA Huntsville, AL* Los Angeles, CA Oakland, CA Ann Arbor, MI Lansing, MI* Colorado Springs, CO Ft. Collins, CO Boulder, CO Denver, CO Atlanta, GA Oakland, CA Dallas, TX Houston, TX Boulder, CO Ft. Collins, CO Co-occurrence Jaccard I Co-occurrence Jaccard I
SoMe Lab Social Medial UW Not significant Linear regression to determine explanatory power of distance for the Jaccard Index of hashtag co-occurrence. Not even close
SoMe Lab Social Medial UW Not quite a conclusion, but we still learned something! Following standard geospatial statistic methods winds up dropping out a lot of data – don’t let the whale impress you. The ephemeral nature of Twitter data poses methodological risks for emergent events User-generated content (in this instance) again follows known patterns of geographic bias Distance is likely more than Cartesian; but distance is also not likely to be solely relative. And the local/global scale remains relevant – GEOGRAPHY MATTERS! (yay!)
SoMe Lab Social Medial UW So…what next? Blockmodeling – attempting to consider multiple similarities between cities: urban/rural, demographics, Jaccard Index, including inputs from SNA analysis from team members Qualitative interviews that attempt to get at some of the reasons folks think they’re chaining related hashtags
SoMe Lab Social Medial UW #Oct20 sukey.io
SoMe Lab Social Medial UW This research was made possible by: NSF Award # INSPIRE: Tools, Models, and Innovation Platforms for Research on Social Media Thank you! Questions and Suggestions?