THROUGH MY WORK with the datasets in this project, and with other datasets in other network analysis projects, I have come to a system of processes for taking data from analogue sources to digital networks. The print-to-network data processing pipeline has the following steps:

• • Plan – Determine the target format or structure.

• • Model – Identify the structure of the data and decide on a data model.

• • Retrieve – Obtain machine-readable text. For printed text, that means:

  • • Scan the print pages into a PDF.

  • • OCR the PDF to create machine-readable text.

• • Segment – Break the text into meaningful units, and assign an identifier to each.

• • Annotate – Mark up each segment to capture the relevant metadata, nodes, and edges contained therein.

• • Reshape – Regularize and reformat the data. Structure it according to the requirements of network analysis tools.

• • Analyze – Explore the data using statistical methods or visualization.

In the rest of this chapter, I dive deep into each of these steps for processing data into a machine-actionable, network visualization-ready format and discuss the considerations required at each step.

Planning

Although analysis is the last step of the process of working with data, it is usually helpful to begin by thinking about the kinds of analysis one hopes to do and the relevant tools available, before embarking on the long journey of data processing—in order to plot a route, one must know the desired destination. This step can be thought of as “gathering requirements”—a process determining the needs and expectations of whatever tools and methods will be employed in the analysis and visualization stage.

Network analysis requires one to identify two basic categories of “thing” in the data: nodes (the people/places/ideas/things of interest) and edges (the relationships that connect those things). To enable interesting analysis and visualization, one needs to know not just what each of these nodes and edges are but what properties they have.