Reflecting Community Events and Social Interactions through Archived Social Media

Webpages and social media associated with geographic communities reflect local events and social interactions. When we archive and analyze these aggregated data of events and interactions over time, we observe a kind of history of a given period for a geographic area, similar to a collection of local information, newspaper stories and citizen commentary. In our data set, we have syndicated (RSS) feeds of information and news posted on websites and social media of local organizations, including government, as well as tweets, blogs and Facebook posts made by organizations and individuals. This kind of collection has built-in biases, of course, just as local print media and online newsgroups do. Nonetheless, the aggregated content reflects a geographic community of users, comprised of individuals as well as organizations, such as, government agencies, businesses, local voluntary associations and residents. We analyzed our data using the standard Latent Dirichlet Allocation (LDA) algorithm and the open source tool NodeXL to identify topics and their changes over time. We also created social graphs based on retweets and @ mentions and quantified exchanges around main topics. Our findings show: 1) distinct topics 2) large and small social interactions around a variety of topics, and 3) patterns suggesting what are called 'community clusters' and 'tight crowd' types of conversations.