For this week’s practicum, I am using Palladio and RAW to show how networks are constructed, what they reveal, and how they can be useful for historians.
I began in Palladio by copying and pasting the battle and unit spreadsheet information into it, and then added in the CSV that contained the battles and coordinates. I first looked at the map view. The map view is interesting since you can see where the battles took place in relation to one another, and you can choose if you want to size the points, which makes the battle sites bigger based on the frequency of their usage in the dataset. The graph view shows the relationships between the units and the battles. If you zoom in closer, you can see that the 44th New York Infantry fought at the Battle of Petersburg and fought at the Battle of Gettysburg, along with the 29th New York Infantry, 1st Michigan Calvary, and the 136th New York Infantry.
Map view with size points
Network showing relationships between battles and infantries
44th New York Infantry fought at Gettysburg along with others
Raw constructs networks slightly differently than Palladio. I began the same way, with copying and pasting the spreadsheet of battle and unit into the interface. I was then asked to choose one chart out of 16 options. Obviously not all charts will work with the data I input. I first tried the Alluvial Diagram.
Circle Packing was another network that worked with the dataset, but it did not present the information in a fashion that I found aesthetically pleasing, nor was it any more or less helpful than the Alluvial Diagram. The Circular Dendrogram presented the information in a way that was easy to look at and can be assessed quickly. The information is not as spread out as it is in the Alluvial Diagram, which I thought made it much easier to read.
I attempted to use the Cluster Dendrogram, but it presented the information in such a way as to render it useless for this particular dataset. Clustered Force Layout was another diagram that was not helpful. Convex Hull, Delaunay Triangulation, Hexagonal Binning, Parallel Coordinates, all required only numbers or dates, so I could not use those. Reingold-Tilford Tree presented the dataset in a way similar to Alluvial Diagram, but it was more organized and easier to read quickly.
My attempts with Scatter Plot, Small Multiples, Steamgraph and Voronoi Tessellation were unsuccessful, but I did have fun playing with Treemap. I was able to get two different results by manipulating the hierarchy and color fields. The first embedded network is with unit in the hierarchy field and battle in the color field, and the second is the reverse.
After installing Gephi about five times, I was not ever successful in getting it to work. While it would install correctly, it would not let me get to the data laboratory.
How are text mining and networks similar or different? With text mining, you are looking at word frequency. This can be used to identify which sources mention the word as well as how often it is used within a corpus. With networks you can examine the relationship between sources. When using Palladio and RAW, I was able to determine the relationship between units and battles, when they overlapped, and which unit fought the most battles. Both of these analytic methods would be considered investigative tools, not communicative tools, since neither of these analytic methods provides any context. Each tool is unique: while we can text mine the content we cannot text mine the networks, and while we can network the relationships we cannot network the content. They can be used in conjunction with one another since they are complementary.
Networks and text mining have taught me how useful digital tools can be in interpreting historical data. They can reveal trends and patterns that otherwise would be buried within the information and they examine the data in ways that humans would never be able to. It is astonishing how we can use technology to further scholarship. I am learning to look at how we can use these tools to analyze sources differently and more efficiently, and how they can be used to further the historical narrative.