The Endo/Exo Writers Project: Computing Class in U.S. and U.K. Novels, 1880-1940
- Principal Investigator
- UIC, Digital Humanities Initiative
Authors and Collaborators: Lennard Davis (UIC), Alexander Dunst (University of Paderborn, Germany), Hannah Huber (UIC), Carla Barger (UIC), Travis Mandell (UIC)
This project aims to analyze the representation of class and poverty in U.S. and U.K. novels by comparing works written by both working-class and middle-class authors. The research builds on an on-going book project by Lennard Davis that describes different patterns of representation based on the economic background of novelists. Thus, narratives that were written by middle-class or upper-class authors frequently show a focus on violence, drug use, urban decay, or other living conditions that is, in comparison, largely absent from proletarian writing. In a first step, a database will compile metadata based on information from about fifty novels by working class (or “endo”) authors and the same number of middle- or upper-class (or “exo”) writers that describe poverty and working-class lives.
This database will be used to construct a digital corpus in a second step. While we envisage that many of these texts will be accessible in digital form, it might become necessary to digitize some of the lesser known works, or those that are still protected by copyright laws. In a third step, we will aim to access larger collections of English works of the nineteenth and twentieth centuries to function as a comparison corpus, either through repositories such as the HathiTrust or those established by other researchers.
Building on working hypotheses developed by Davis, we will then seek to establish whether there are quantitative differences between “endo” and “exo” representations of class. Several established methodologies within digital literary studies may be drawn on for this project, including natural language processing (NLP), social network analysis, topic modeling, and sentiment analysis. NLP methods will show whether these groups of texts differ on a basic word level, including the use of key words, collocations, or most frequent words. NLP may also serve as an indicator for the presence of non-standard sociolects used by working-class characters and the use of certain adjectives or verbs associated with them. These findings will shed light on the description of living conditions and character agency within these works. Social network analysis has the potential to reveal differing representations of the social milieus of working-class characters, their integration into family units or participation in larger networks (i.e. labor collectives or ethnic neighborhoods). At a more complex level, word embeddings (Word2Vec) may highlight associative clusters that establish the wider context of key words. Similarly, so-called topic models allow for the identification of narrative themes running through corpora of texts. Finally, while sentiment analysis currently works best for shorter, non-literary texts, for which it was developed, it might be worth exploring whether working and middle-class characters in the corpus are associated with positive or negative emotions.
Abstract by Alexander Dunst.