Dialogism in the novel: A computational model of the dialogic nature of narration and quotations

Felix Muzny, Mark Algee-Hewitt, Dan Jurafsky

Abstract:

Understanding how spoken language is represented in novels over time is a key
question in the Digital Humanities. We propose a new metric for characterizing
spoken dialogue in the novel, called dialogism, that instantiates Bakhtin’s claim that
all texts are fundamentally dialogic. This measurement uses abstract grammatical
features in a span of text (such as the use of pronouns, mood, or subordinate clause
structure) to measure the extent to which the span is dialogic, i.e. exhibits the
grammatical structures common to natural spoken dialogue. We use this metric
to explore the dialogism of 1,100 largely canonical English novels over 230 years. We
combine quantitative and qualitative investigation of the dialogic properties of both
dialogue and narration to show novel stylistic properties of literary innovation
during three periods: the late 18th century, the turn of the 19th century, and the
mid-20th century. We find that during these moments, certain authors reject liter-
ary conventions by changing the dynamic between the narrative and dialogue por-
tions of the texts. Our analysis shows that these changed dynamics are behind rises
in persuasive writing, reflections of psychological processes, and the use of dialogue
as an increasingly important driver of the novel as a whole. These results show that
computational models that characterize style grammatically, generalizing across
time and genre, can lead to literary and methodological insights.