Sentiment Analysis of Late Victorian Novels: Opportunities and Challenges

Abstract

Computational research has been steadily gaining ground in literary studies and Victorian fiction has been one of its main objects of study. However, this research has mostly sought to empirically corroborate concepts proposed by literary scholars and has been out of sync with theoretical considerations from the field of cultural evolution and the cognitive sciences. This paper aims to show the massive potential of this theoretical framework for guiding empirical research in literary studies, by applying it in a hypothesis-driven sentiment analysis of late Victorian novels. With the help of the “syuzhet” and “sentimentr” packages in the R programming language, this paper analyses the emotional valence and emotional arousal of a sample of 846 late Victorian novels published in the UK, available on Project Gutenberg. The main hypothesis of the paper is informed by the cognitive sciences-novels with a lower emotional valence and higher emotional arousal would be more successful in cultural transmission, which would in turn increase their cultural longevity. The analysis also looks into the relationship between the derived sentiment values, the genre of the novels and the gender of their authors. Finally, the paper deals with the limitations and potential flaws of such analyses and outlines several possible trajectories that future investigations can take.

Stefan Veleski

Doctoral student in English literature