Main information - Project Outline - Works Consulted - Ways of Dividing Interactions - Preliminary network of "The Teacher" - Pre-pilot Survey
Using social network analysis to explore narrative structure
'''A TRANSIT 2011 project.'''
Student researchers:
Andrzej Jarynowski (andrzej.jarynowski[at]yccsa.org), Stephanie Boland (stephanie.boland[at]yccsa.org)
Supervisors:
Dr. Elva Robinson (YCCSA), Dr. Dan Franks (YCCSA), Dr.John Forrester (YCCSA), Dr. Richard Walsh (English).
A social network is the map of interactions between individuals. We use social network analysis to understand and explore social structures, such as whether the friend of my friend is also my friend – or whether the enemy of my friend is also my enemy. Social network analysis has been applied to human societies by sociologists and to animal societies by biologists for many years. In this project, we want to make a cross- disciplinary leap, and use the tools of social network analysis to explore social structure within fiction. We could investigate such questions as: Can we characterise the genre or period of a book from aspects of the social network it represents? More specifically, do the social networks described by authors writing about well-functioning societies differ in recognisable ways from those writing about dystopia? Does social network analysis
provides ways to articulate plot developments in novels? To what extent does social network analysis correlate with the interpretative activities of different readers, for example, do different readers infer different social networks from the same novel? The results of this project will provide new interdisciplinary insights, and will demonstrate to what extent social network analysis can be usefully applied to the study of narrative.
From the initial project description: (http://www.york.ac.uk/res/yccsa/research/transit/scholarships2011/201103.pdf)
The project, working with Sherwood Anderson's short story series ''Winesburg, Ohio'', focuses on how social networks are read in literary texts. By comparing networks produced by automated systems with those created by surveying more subjective human responses, the project aims to investigate the different ways social interactions are 'read' in texts.
* Elson, David K, Dames, Nicholas, McKeown, Kathleen R. 'Extracting Social Networks From Literary Fiction. (http://www1.cs.columbia.edu/~delson/pubs/ACL2010-ElsonDamesMcKeown.pdf)
**Recent research from Columbia, extracting social networks from a wide range of texts through automated methods. Compares the different social networks in first and third person narratives, and in the urban and rural novel. Uses 'character name chunking, quoted speech attribution and conversation detection given the set of quotes' (1).
*Phillips, William L. ‘How Sherwood Anderson Wrote Winesburg, Ohio’. American Literature, Vol. 23, No. 1 (Mar., 1951). North Carolina: Duke University Press. pp. 7-30
**Exploration of the biographical and genetic (i.e. structural) process behind the writing of Winesburg, Ohio. Does not deal specifically with the social networks, but reveals something of how character relationships grew. Gives an interesting insight into the editing process.
n.b. ''This list is far from complete, and should be read as an 'edit in progress'!''
*Hanneman, Robert A. and Mark Riddle. 'Introduction to social network methods'. Riverside, CA: University of California, 2005. (http://faculty.ucr.edu/~hanneman/)
**A useful web resource, giving an introduction to social network methods. Proven understandable, even to a humanities student.
Ways of Dividing Interactions:
Early on in the project, several ways of dividing social interactions were posited; interactions that were 'implicit' and 'explicit', interactions that were primarily 'thematic' or 'plot-driven', etc. After discussion, it was decided that the only distinction that was to be made in the surveys was between interactions that are 'important' and 'unimportant', with the reader left to decide what they considered important ('your reading of the story'). This was then further developed to allow the reader the ability to rate each interaction out of ten, giving, hopefully, a more nuanced result.
One important distinction noted by Dr.Walsh that must be paid attention to during later analysis is that between actions important within the world of the story, and actions that form an important part of the reading experience. This distinction must be considered during both interpreting the human survey results and in developing a method by which to automatically 'read'.
The models below show when the distinction between 'implicit' and 'explicit' interactions was still in place.
Preliminary network of "The Teacher":
This network was generated by one person's reading of the short story 'The Teacher'.
In this context, an 'explicit' interaction is one which is shown in the text, i.e. two people speaking or touching. An 'implicit' action would perhaps be better described as an 'indirect' one; it is something referred to or remembered, rather than shown to the reader. Although this distinction permits an interesting investigation into whether or not interactions we 'see' are judged as more important, it was dropped from the full pilot survey due to ambiguity.
Fig. 1. Implicit (left) vs Explicit (right) interactions
These models show the difference between interactions noted while going through the text, and the overall importance of the relationships upon completion of the story. Six people took part in this pre-pilot: students and supervisors.
Fig. 2. Interactions noticed (left), Overall view (right)