The project allowed me to explore many different sources of data visualizations. Rather than attempting to find a new meaning to Only Revolutions, I analyzed the issues with information visualizations. Being the non-tech-savvy person that I am, I stuck mainly to Voyant, TaPoR, and Dipity. These programs alone provided me with ample reasons that machine reading cannot provide an encompassing representation of a literary text. In order to refine the focus of my paper, I chose to explore the issues associated with text anaylsis, the physical visualizations, and the biased nature of reading (versus the concrete nature of data). The following three visualizations formed the foundation of my arguments.
This visualization is an assortment of the most commonly occurring words in the narrative of Only Revolutions. This visualization demonstrates multiple issues in one image. First, it is very colorful and complex in its presentation. Observers will see the pretty colors and become enthused. Once they are interested, users ignore the need to analyze the validity of the visualization. They fail to ask the question, “Is this an accurate depiction of the text?” which leads into the next problem with this visualization. The second problem is that visualizations don’t necessarily make accurate claims about a text. These words are the most commonly occurring words in the novel, yet they are all very generic words used to allow a sentence to make sense. Words such as “and,” “the,” and “with” clearly have no true importance in Only Revolutions, so this visualization is essentially useless.
The second visualization is a graph showing the frequency of the word “kiss.” The point that this visualization demonstrates is the limitations of human perception. There are 10 specified data points along this graph, but the rest are simply not given an exact value. The only way to find the rest of the values is with an estimate. Can you tell me what the frequency of “kiss” is (rounded to four decimal places) when x is 5? No, you cant. The human eye doesn’t have the capability to predict that value without a large margin of error. If you can’t accurately define data points, then how could you expect to make conclusions based on that data?
The final visualization that I decided to present is a DocuBurst (from TaPoR) that shows the occurrence of words liked to “sex” and their frequencies (denoted by the size of their portion of the circle). This was an attempt to overcome the issue with text analysis in which the difference in frequencies between “kiss” and “sex” could lead to varying conclusions. It seemed to resolve that problem because “kissing” was a word that appears on this visualization. However, there are still phrases that suggest sexual activity that don’t appear on this visualization (ex. “gobble my nob” (202H)). The point being made is that even when we think we have adjusted for error, we can still miss details and phrases that a data visualization does not have the capacity/intelligence to provide.