Dada Networks

Summer workshop report, 2016

[Presentation slides in PDF]

The Dada Networks project, led by principal investigator Kurt Beals, has been using journal publication networks to explore the structural contours of the Dada movement. The project has predominantly used the software application Gephi to create and visualize its networks—a free, open-source tool commonly used in the digital humanities. This summer, however, graduate student Melanie Walsh (with help from graduate student John Ladd) has also explored some other methods (NetworkX and D3.js) for creating, calculating, and visualizing these networks, which throw into relief the strengths and weaknesses of Gephi, as well as how these tools bridge humanities scholars to the fields of computer science and statistics more broadly.

The Python software package “NetworkX,” for example, requires more technical programming knowledge than Gephi right off the bat, as evidenced by a simple comparison between Gephi’s more intuitive Photoshop-style interface, where networks can be manipulated live, and the lines of Python code. But NetworkX offers a number of features that Gephi does not, including the capability for calculating bimodal network metrics, the capability for taking nodes and edges straight from the main database spreadsheet (as opposed to being manually re-manipulated into another spreadsheet format compatible with Gephi), and the capability to output the network into different kinds of files, including Gephi files (making NetworkX and Gephi not mutually exclusive tools). NetworkX can also output the network into a .json file, which can be used to create an interactive web-visualization through the JavaScript library D3. The shareability and interactivity of this web-based network visualization offers great pedagogical potential and a more intuitive sense of how the fore-directed algorithm works (e.g., nodes are not geographically rooted in space), as well as the ability for other scholars to explore the same dataset to cross-validate conclusions or investigate other research questions entirely. There are many features that can be added to the D3 web visualization that would allow a potential user to explore and manipulate the network, including but not limited to: a search box that can locate any node, the ability to click on a node and highlight its immediate neighbors, and a sliding scale that can remove nodes below a certain degree threshold. (For an example of the wide range of potential features, see this D3 network visualization by Elijah Meeks and Maya Krishnan.) In conclusion, NetworkX and D3, though more technically demanding tools, seem like promising ones for creating, calculating, and visualizing networks, which can be used in isolation from or in tandem with Gephi.