Crowd Sensing System for Public Participation

GIScience 2016 Short Paper Proceedings Crowd Sensing System for Public Participation M. Tenney 1 , R. Sieber 1 , G.B. Hall 2 1 McGill University, Montreal, QC, Canada Email: matthew.tenney@mail.mcgill.ca, renee.sieber@mcgill.ca 2 Esri Canada, Toronto, ON, Canada Email: bhall@esri.ca Abstract We propose a crowd sensing system to capture certain dynamics of public participation in a city. Crowd sensing systems (CSS) attempt to capture the opinions of local publics from web- resources. We define our CSS using a spatially-situated social network graph where users along with different variables, such as time, location, social interaction, service usage, and human activities can be studied and used to identify experts or influential citizens who are relevant to municipal affairs. 1. Introduction In the fields of engineering and the computational sciences, the term crowd sensing represents a popular area of research (Cardone et al. 2013). Similar to citizen sensing, urban sensing and participatory sensing, we broadly define crowd sensing systems (CSS) as being an integrated hardware and software architecture designed to collect user-generated content for a specified topic, issue or theme. In this paper, we introduce a portion of our conceptual CSS, which describes several social and spatial interactions within a local population (i.e., connections between individuals and locations of communities-of-interest), establishes place- based topics across a city from user-generated content (e.g., geotagged posts from social media), and identifies various forms of activity across specific geographies (e.g., patterns of urban travel). The CSS combines methods of natural language processing, spatial analysis, and graph theory to create a data structure with possible value when used to inform local decision makers. Our work builds on smart city initiatives, data-science and Web 2.0 literature that seek to revise traditional forms of public participation (Cardone et al. 2013). In particular, these difficulties can be assuaged by integrating data-driven techniques that automatically extract “similar” information (i.e., topical pubic opinions) from user-generated content. 2. Crowd Sensing Systems as Tripartite Network Public participation in municipal affairs is often seen as a product of stakeholders and interest groups, which are spatially distributed across a city. Choosing to model public participation digitally, requires representing relationships between structured and unstructured content, deriving explicit and implicit social interactions, and inferring frames of context through shared interests and co-location. Like other social networks, our CSS network graph (G) contains nodes and edges G = (N, E). The graph is further divided into three subgraphs containing unique node and edge types defined as U = (Un, Ue), C = (Cn, Ce) and T = (Tn, Te), where Un, Ue are nodes and edges of the user profiles in subgraph U; user-generated content forms the subgraph C consisting of nodes Cn and edges Ce; and a “geotopics” subgraph T contains spatially located nodes Tn with temporally weighed edges Te. These connected subgraphs as shown in Figure 1, can form spatially-situated networks constructed from what we call Social Signals (SocSigs, see below). SocSigs information contained in each of the connected sub-graphs include a user-network U (i.e., social-network among