A computer’s understanding of literature

Everyone agrees that a computer could never understand and appreciate literature, but the fields of computational linguistics and discourse processing have made important advances in automatic detection of language and discourse characteristics. We have analyzed literary texts and political speeches with two computer tools, namely Coh-Metrix and Linguistic Inquiry Word Count (LIWC). Coh-Metrix provides hundreds of measures that funnel into 5 principal components: word concreteness, syntactic simplicity, referential cohesion, deep cohesion, and narrativity. LIWC classifies words on 80 categories, such as first person pronouns, negative emotions, and social words. This paper illustrates how computer tools can unveil new insights about literature and can empirically test claims by literary scholars and social scientists. Our approach offers a computational science of literature.