To many folks interested in the geography of Twitter, journal articles like this or this are representative of nascent research in the field. Typically authored by folks outside of geography, articles like these correlate a person’s location with their relevant Twitter activity. However, geographers have resisted the notion of space and place being restricted to Cartesian notions of “where” to include the socio-historical context of situated, lived experiences in the city.
My work, as illustrated here and here, considerably troubles these waters by considering Twitter place-based interest networks as places unto themselves. Certeau reminds us in The Practice of Everyday Living that the becoming of a city consists in large part of the banal, everyday exchanges that form our everyday practices of living. Certeau tells us to “walk the city” to gain information about it; I argue too that one must “walk the data” to gain an appreciation for its relation to place. These data are everyday exchanges -- sometimes banal, sometimes much less so -- that comprise, at least in part, the relations between individuals and the places about which and through which they tweet.
I take over a million tweets across a year’s worth of Twitter data regarding Occupy Wall Street, and treat it in three ways to illustrate the ways in which one might computationally “get at” place without being beholden to Cartesian mapping. I use social network analysis to lay out the topography of a network -- where users are connected, those users that are most visible, and users that group together into place-framing clusters of activity. I use Bayseian topic modeling (LDA) to perform what Ramsey refers to as a “distant reading” of texts, allowing me to sacrifice the granularity of a “close read” for the ability to gain a course understanding over a massive set of texts otherwise unreadable in this lifetime. And finally, I use qualitative semi-structured interviews to get at the parts of Twitter exchanges that are otherwise invisible through the APIs we use to collect them. All of these tell a story about a communicative place-based information network that move beyond “the map” as singular to reveal multiple mappings, and multiple readings of those maps.
And unlike most big data research, I situate myself as participant-observer -- not as a detached God’s eye view from cyberspace, but as an active constituent of the place-based networks which I study. Few researchers can advanced a situated view of big data at least in part due to the seduction of the methods being used. But as an example, I once came across a tweet that said “Occupy Disneyland!” and featured an image of two children on a water slide. Naïve use of the tools above might suggest some activity at Disneyland related to protest; while I, as a situated researcher (i.e., “snoop”) discovered that the user was a sarcastic banker whose other tweets fell largely into the “dirty hippies need a job” category.
Walking the data requires more than plotting the points contained within metadata, tracking exchanges of information against transportation patterns, or geocoding free-entry text fields that supposedly relate to a person’s location but can just as easily list “The Moon.” It requires, as Wilson writes, a consideration of Twitter as a phenomena unto itself, and not necessarily reflective of existing phenomena. While this can utilize a wealth of computational tools, it behooves us to consider geography as something-more-than a set of latitude-longitude coordinates -- a set of situated, embedded practices that allow a user to relate to and shape places, both online and off.