We are currently at a stage of academic research in which the ‘statistical analysis of literature has gone from crackpot theorising to cutting edge research.’ As I discussed in a previous post about Franco Moretti’s distant reading concept, the way in which English Literature can be approached and analysed is rapidly expanding. After initially being sceptical of other research methodologies the more research I do on the internet or in the library, the more convinced I am that other research methods should complement close textual analysis. As part of the set reading for a seminar we had to read Lev Manovich’s article ‘How to Compare One Images’ in which he presents the research methodology ‘Cultural Analytics’.
In his article Manovich describes cultural analytics as ‘the use of visualization to explore large sets of images and video. These visualizations can use existing metadata […] and also new metadata added by researchers via annotation or coding. Manovich’s concept is similar to Moretti’s distant reading as it uses statistical analysis and computer systems to investigate and explore large sets of big data. The primary contrast between cultural analytics and distant reading is Moretti’s concept analyses large sets of textual data whereas Manovich’s concept examines large sets of images and videos. As with Moretti’s distant reading, cultural analytics was a research methodology I was unaware of until I encountered it on this module. One question which is continuously encroaching on my academic research is ‘How does the notion of scale affect humanities and social science research?’ The purpose of this post is to illustrate the strengths and weaknesses of working with big data using the cultural analytics methodology.
So let’s start with the strengths of Manovich’s concept and the opportunities it can offer an academic who is using this particular research methodology. Similar to Moretti’s distant reading using cultural analytics is a useful method to use if you are working to a deadline as it can provide a large amount of data in a short space of time. Manovich argues that humans do not have the natural visual system to notice subtle visual differences in images neither do we have an adequate vocabulary or metalanguage to describe these visual differences. However Manovich proposes by using digital image processing and computers which are capable of ‘feature extraction’ academic researchers can measure anything from presence of texture to number of edges on an image. So not only can cultural analytics provide a large amount of data in a short space of time, it can also provide data which is so in depth and detailed it exceeds human visual capabilities.
A significant part of my project is thinking about how I can represent a large amount of textual analysis in a visual format. In Morris Eaves’ article ‘Picture Problems: X-Editing Images 1992-2010’ he discusses the limitations of human based research into describing the details of large sets of big data. These limitations are illustrated by Eves through his discussion of the William Blake archive. Eves argues the central issue with human based research and analysis is a lack of awareness of methodical boundaries. For example if I was describing an image from the William Blake Achieve, how much detail do I need to include in the descriptions and tags to offer adequate information for other academics and researchers? Do I have the necessary vocabulary or metalanguage to describe the details of the image? Manovich’s article examines how the limitations of human capabilities which Eaves outlines can be counteracted by a strict methodology and suitable technology. Manovich uses an image set from the Japanese comics Manga to illustrate the importance of a strict research method and an accessible digital technology. Manovich uses a scatter plot graph with two axes, an X-axis to represent the standard deviation to measure variability. The second axis is the Y-axis which represents the entropy over greyscale values. The digital scatter plot illustrates that by using a strict methodology there is no chance of losing track of your original methodology or thesis (a common problem in academic research).
Spontaneous discovery is one of the stand out strengths of cultural analytics because it allows for unprompted discovery of interesting patterns in an image set. Manovich argues spontaneous discovery of patterns isn’t possible with other consumer software and web services such as slide show because images can only be sorted by a few parameters such as uploaded date or file name. In the final part of the article Manovich discusses human defamliarisation with computers which I interpreted as a massive strength within his concept of cultural analytic. The ability to analyse a set of images along a singular visual dimension allows researchers to see what they had not noticed during previous analysis. We can use cultural analytics to ‘defamiliarize our perceptions of visual media cultures.’
Despite the strengths of cultural analytics, I will not be using Manovich’s concept in my project because I am working with texts rather than images. Another reason for avoiding cultural analytics as a research methodology is because I am still sceptical as to how accessible and understandable it is for both academics and non-academics.
[4.] Available at http://digitalhumanities.org/dhq/vol/3/3/000052/000052.html