Detecting Keywords: Planet and Narration project
August 22, 2014 11:23 am
Leave your thoughts
The Planet and Narration research project investigates depictions of “survival” that appear in the Brundtland Commission Public Hearings. Drawing from philosopher Giorgio Agamben’s study of “bare life”1, this project is particularly interested in discussions of “basic need” and “bare-life” survival as they relate to representations of the Global South. For Agamben, to be reduced to “bare life” is in some sense to no longer be human, and to be outside of political community altogether. The digital humanities Voyant tools were used to identify keywords related to “survival” in the transcripts, including “life,” “death” and “basic need.” The tools helped us to sift through the cascade of over 35 documents comprising the hearings, many containing over 100 pages of text, to explore how conceptions of “bare-life” survival are framed in the hearings.
I wanted to create a situation where words and themes would emerge through the tools which I would not have thought to explore otherwise. I began with the terms “life,” “death,” “need,” and “survival” and used Voyant’s Document Type KWICS Grid tool, also known as the Keywords in Context tool, and Links tool to try and discover other terms related to these themes.
Links is a Digital Humanities tool that represents words that reoccur in the same context as others through a network of nodes and edges. Basically, when you search for a term, it presents that term with a small collection of other terms branching off from it that possess the strongest level of adjacency (i.e., they occur most frequently in the same context as the term). You can increase the number of terms branching from the word you searched by clicking that word’s node. The more you click, the less adjacent the newly-generated words are to the word under study.
In addition to using the Links tool to identify key words that might escape a researcher engaging in a straightforward read through of the hearings, I also used Keywords in Context, a tool that contextualizes terms by generating a table containing the phrases that come before and directly after a searched word. This was useful for quickly scanning for significant terms related to “survival” that may not appear in the context of that word very often. For example, when reading through the table generated from the word “survival,” I was excited to come across the word “existence.” It is not surprising that “existence” did not appear very often in the same context as “survival” since it is effectively a synonym that serves as a substitute for “survival.” While this promising word remained undetected through the Links tool, the one or two instances in which it emerged in the table generated through Keywords in Context allowed me to find it.
As I encountered interesting terms, I ran them through the same tools that brought them to my attention in the first place. I found that examining different words related to “survival” brought different discourses to the surface. Plotting “survival” in the Links tool, for instance, generated “human,” “indigenous,” “life,” “man,” “poor,” “poverty,” and “threats.”
But while “survival” appears to be used in discourses around the future of indigenous people groups and of people living in “poverty,” plotting the word “existence” in the Links tool suggested an entirely separate set of contexts.
“Species,” “forest,” and “scientists” signalled that “existence” is used in discourses of species survival, while “soviet” suggests survival in the context of the nuclear arms race of the Cold War. These tools thus helped us identify how survival has multiple meanings in the environmental discourses of the hearings, and that these meanings are specific to particular social and geopolitical contexts.
We also made some observations about the limits of the tools. It became abundantly clear to us that terms appearing far more frequently than others in the corpus were not necessarily worth pursuing or investigating. The size of the nodes in the Links tool denotes the amount of times the word appears in the corpus. As you can see in the pictures above, “environment” and “development,” on the basis of their size alone, appear at first glance to have a significant relationship to the terms “survival” and “existence.” But when you consider that both terms appear in the name of the commission superintending the hearings —the World Commission on Environment and Development — and that the commission is routinely mentioned throughout the proceedings, it becomes much harder to read these terms in meaningful ways. A purely quantitative approach to Digital Humanities might have overlooked the smaller nodes in the graph that have become such a rich source of discussion for this research project.
The Voyant tools can be used to efficiently detect significant keywords and themes that might escape a researcher who is trying to interrogate large bodies of text. This post has outlined how the Links tool and the Keywords in Context tool have been particularly useful for tracing important words that construct imaginings of survival in the Brundtland hearings. In the next post, I will describe how we are making further use of the Keywords in Context tools as we continue studying these documents.
1. Agamben, Giorgio. Homo Sacer: Sovereign Power and Bare Life. Stanford: Stanford UP, 1998. Print. p. 8