While embarking on my journey into the land of Digital Humanities, my course on Digital Public Humanities (DPH) has ensured that I am not low on provisions. Yet another tool I have explored in this course to make sense of the data and information spread throughout the digital landscape is one concerned with networks and visualizations: Palladio! This program is provided by Standford University and Humanities + Design to help users (and students) plot data with ease and create visualizations of said data in the form of networks and graphs.
As with the previous posts, I will offer a brief guide on the functionality of this program while using my assignment as an example of how the options can affect the representation of data. At the end, I will record some thoughts on what I learned from using this tool.
Accessing Palladio and constructing your visualizations begins with entering your data. After hitting the “start” button on the front page, you will be taken to a blank page with a text box. This is where you will drop an (appropriate type) files that contain your data. Do this and hit the “load” button to initiate the rest of the options. You will soon see six options in the top left corner. Under the “Data” section, this is where you will provide the rest of the information necessary to create layers (for the map function) or nodes (for the graph).
In this section, you can see how the data has been defined by the metadata of the files you provided. Clicking into these options will reveal options to add tables and layers. These options are for adding more data that you can use to interact with the first dataset you entered. For example, I first uploaded the general interview information of former slaves from the United States. This contained biographical information of the interviewees and details specific to what was spoken about during the interview.
Using the layering options, I also added geographical information for where the interviewees were originally enslaved and where the interviews they did were conducted. This informed the options for the map functionality. Going to the “Map” tab will present a similar looking map as seen in the previous post about kepler.gl. From here, we select the kind of marker we want displayed–I chose point-to-point, for this will present the data similarly to how it appeared on my kepler.gl model. Then selecting the source option, I inserted the location for the interview and the location for enslavement in the target option. This then plotted my data onto the map.
When switching to the graph, these same datasets are used to build the graph. The sidebar menu on the far left is how you can select information. Select “source” and pick a category to display that information as nodes on the page (this forms the first part of your network). Select the “target” option and pick another category. This is the second node option that will create connections between it and the first node, connections based on the relationship between those two items. For example, selecting “interviewer” for the source and “interviewee” for the target will then diagram with cluster markers and line connections the network of which interviewers interviewed which interviewees (this sentence was on purpose). Clicking the “Highlight” option will better define the difference between the nodes and the “Size nodes” also better distinguish between the different points of the dataset. In general, the larger the cluster, the more points of data it includes (pertains mostly to when there is distributive factors in the data being displayed, such grouping female/male interviewees together would present two large clusters of interviewees).
Speaking of grouping, the bottom ribbon allows you to apply filters. Opening this up (by clicking “facet”) will offer the same categories available for defining the source and target data. Selecting “dimensions” on the far left and choosing the categories will add them to the drop down menu that has now appeared and then selecting the options that appear there will allow you to refine the nodes appear in the diagram. For example, if I chose the dimension of interviewers, I could then filter out specific interviewers from the nodes in the diagram.
When you have the desired data diagrammed, you can click and grab nodes to rearrange the plots as you see fit. To export an image of the diagram, click the download button on the far left.
What I Learned
This program really hit home both the benefits and the problematic nature of this kind of tool for Humanists. This type of tool can be of massive help when trying to plot the relational aspect of information and the connection we are driven to see when even peering into the cold world of hard numbers. Because more qualitative characteristics could be worked into the algorithm, the data could be defined in a way that can benefit humanities research. For example, selecting gender for the “Source” option on the graph and “Topics” for the target created a network that depicted two large clusters–one for female interviewees and one for male interviewees (as they were the only ones who would have been defined by gender in the initial input file)–and many small cluster markers in between them to show what topics were discussed by both female and male interviewees while being flanked by fewer small clusters associated only with the female or male cluster, indicating those topics were recorded as being discussed only by that gender. This is useful information because it raises questions about that facet, allowing for analysis, asking questions, and interpretation. What topics were defined by gender roles in the 1930s by former slaves? What topics did they share in common? Why were those topics discussed across the gender spectrum, but not others? Which gendered group was more open to discussion on the recorded topics? What circumstances led to them only discussing certain things and refraining from others? These are some questions that can be asked with the data plotted onto the graph.
Yet, the drawbacks hit the user fast. When it comes to mapping geographical information, it is clear that mapping tools do this more effectively. The diagram for the graph feature lacks a map makes geographical filters and parameters more abstract and not as useful. Additionally, being unaware of how the data should be define can be detrimental because of the very reason this tool is useful: relational connections. The nodes are having relations drawn to other nodes via the algorithm. Therefore, if the source (or target) data isn’t categorized in a way that limits the amount of data being drawn upon, it can create a large volume of nodes and this will create an excess of connections (also called “edges,” forgot to mention that earlier) and clutter the diagram so much that the data becomes virtually distorted and unintelligible. This is what we call “dense,” as opposed to being “sparse” when there are few nodes/edges that make the diagram appear cleaner.
This assignment has definitely stressed two things for me: 1.) try to pick the right tool for the job (or at least recognize that you don’t have to use one tool to solve all problems) and 2.) learn what your tools are capable of doing with the data you have at hand. Networking tools are cool and can make informational visualization, but they won’t be as fun or informational if you’re trying to plot data that is better suited to other tools.