5. Twitter Bootstrap and Responsive Design
1.3 Quantitative/Scientific v Qualitative/Humanistic?
In his book, Reading Machines, Ramsay writes about the meeting between scientific and interpretative methods in literary analysis and argues that digital text analysis needs a framework for operation which admits both these modes of inquiry. He argues that instead of adopting a purely scientific approach to text analysis, in that a hypothesis is proved or disproved through evidence gathered from an analysis of data, there should be an aspect of interpretation involved. Text analysis should not attempt to prove an interpretation of a text but should instead aim at intervening in discussions about a text.
Topic modelling itself, for example, relies on interpretation. The topics produced by the topic modelling program are simply collections of words that occur frequently throughout the corpus. It is up to the scholar to provide a label for these topics. This is a highly interpretative act and provides space for debate.
In the meeting between the humanities and the scientific method, the humanities is often framed as the side that must compromise its own methods in order to benefit from a more scientific analysis of literary data. But Ramsay argues that the scientific method, when meeting with the humanistic, must also compromise and admit the hermeneutical method of traditional literary studies. Text analysis must not set out to “prove” hypotheses but instead must offer potential readings of texts that intervene in ongoing discussions. Digital text analysis is used in this edition of Melmoth the
Wanderer, for example, by supplying users with visualisations of topic models in order
to offer important context in which to view the core text. These topic models do not claim to be definitive representations of Irish writing or gothic literature around the time Melmoth the Wanderer was written but instead are potential readings based on the parameters and data used. The interpretations of these models offered in the paratexts are also subjective and are up for debate. It is important in this regard that results from text analysis allow for critique from both scholars using similar methods and from scholars using traditional interpretative approaches. If a topic model is shown not to apply to a particular novel from this period as a result of a close reading carried out by another scholar then that must be acknowledged and the entire topic model itself must be questioned. Just because a topic models has been generated using more scientific methods that are involved in a close reading does not mean that the topic model has a greater claim to being correct than the close reading. Failure in distant reading and in
processes like topic modelling must be allowed. Ramsay, in fact, sees failure in textual analysis as an opportunity. He asks if such analysis would be better off eschewing success and instead asks “how much more gloriously or fruitfully it might fail. The goal, after all, is not to arrive at the truth, as science strives to do. In literary studies, as in the humanities more generally, the goal has always been to arrive at the question” (68).
Much of the criticism of quantitative analysis of literature suggest that its method is too rigorous and too scientific. Bode believes that such criticisms need to be engaged with if quantitative methods are to “make a productive contribution to literary history and humanities scholarship” (7). Bode suggests that two of the primary criticisms of quantitive analysis and of Moretti’s work is that it reduces the complexity and multiplicity of literature and language and that it makes false claims to authoritative and objective knowledge. Critics suggest that it is an inappropriate methodology that flattens out the field of literary history and loses the specific and particular details that close reading can uncover, while attempting to negate the results of close reading through the championing of distant reading’s more “scientific” methodology. For critics of quantitative approaches to literary studies, the rhetoric that accompanies these approaches is harmful for the humanities in general. As Bode argues, “numbers and statistics are imbued with significant power in modern society” and “much of this power comes from the rhetoric of objectivity and truth surrounding” quantitive analysis” (11). Such rhetoric is also employed in the use of computational methods, and this, for Bode, leads to a channelling of funds towards projects that use computational and quantitative methods. She believes that such a configuration of knowledge will have “major negative consequences for the humanities” (11).
But these two methods do not have to be an either/or choice. Bode points to the “methodological paradigm” of book history as a helpful example of how qualitative and quantitative methods can inform each other. In book history, instead of these two methods being divided and opposed, they are seen as two different perspectives with different uses and applications to the study of the literary field. Bode cites Joshi, who shows how close reading can reveal things that numbers and statistics cannot and vice versa, and Zwicker, who sees the loss of detail that comes with quantitative methods as justifiable when such methods contribute to questions that remain unanswerable using qualitative methods (14). Bode uses the example of book history to show that, rather than fostering an opposition between these two methods, the study of literature benefits when the advantages of both these methods are acknowledged and are used to inform each other. Neither qualitative nor quantitative methods can tell us everything about the literary field but, when taken together, their perspectives can afford us a better view of the field as a whole. Digital critical editions, as tools of close reading and as located in the digital space, are in a privileged position to become a potential meeting point between these two methods. Gabler imagines the digital scholarly edition as an interrelated web of discourses (“Theorizing” 44). The discourses or paratexts of this edition of Melmoth the Wanderer use topic modelling, for example, to offer users context in which the core text can be viewed. These include network visualisations of corpora of both Irish novels and gothic novels published during the late-eighteenth and early-nineteenth centuries. The data from these topic models is also made available to users. While browsing the text, users can choose to view this data and organise the text according to the terms found in the topic model. This allows users to move from the general results of the topic model to the specifics of Melmoth the Wanderer. In adding
this functionality to the edition, users can not only browse the edition in a unique way but they can also interrogate the topic model. By making the results of the topic model available in this way, users can bridge the gap between the qualitative and the quantitative and see how the topic model actually applies to the text of Melmoth the
Wanderer. They do not have to accept the results of the model without question but can
instead see where each of the terms returned by the model occurs in Melmoth the
Wanderer. They can then discover for themselves if the patterns suggested by the model
actually do exist in the novel or if there are in fact errors in the model.