During the Esri Federal GIS conference in Washington DC this week, the analysis of social media data was the topic of several presentations and demonstrations on the exhibit floor.
Andrew Turner and Sean Gorman, formerly of GeoIQ which was acquired by Esri, gave a presentation on how data derived from social media can be aggregated to provide relevant information in near real-time. They demonstrated how by simply showing the location of geotagged flickr photos (estimated now to be over 1 billion) that new types of geographic boundaries can be discerned, such as those areas that may be relevant to tourists versus that which may be significant to locals. Using the search API from Twitter, they looked at other trends but cautioned that using this alone may not be giving you a random sample in order to qualify something, such as consumer sentiment, as a "trend."
However, using the streaming API might allow for the identification of real-time events. They demonstrated how a user can analyze data within certaiin temporal and spatial constraints. As such, it is possible to set alerts based on parameters set by the user. One of the goals that Gorman and Turner are trying to reach is to detect patterns without having an analysts constantly monitoring the streaming data. Their objective is to have social media patterns reveal themselves. They want to be able to look at all the conversations on Twitter and find the relevant conversations on a specific topic during a particular time period. Would a certain keyword show up more than normal and thus reveal an anomalous event? If so, they are looking at ways to set alert to send text messages to analysts.
Their team at Esri is looking to work with companies such as Topsy
to better discern the validity of streaming social media. Topsy was an exhibitor at the conference.
Another exhibitor, TerraGo Technologies was demonstrating a recently announced
application for analyzing unstructured data. Using their GeoXray
product the company will geographically display the locations of blogs, news feeds, Twitter, etc. and filter by place, topic or temporal constraints. The interface can use Bing or Google maps. The objective, according to Rick Cobb, CEO, is to provide better situational awareness to its clients. It has applications in defense but could also be applied to commercial uses for companies who likewise need better visibility regarding situations that may result from natural disasters in order to support merchandising and logistics. (Image below courtesy of TerraGo)