Casting Light on Invisible Cities: Computationally Engaging with Literary Criticism

Proceedings of NAACL-HLT 2019, pages 1291–1297 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 1291 Casting Light on Invisible Cities: Computationally Engaging with Literary Criticism Shufan Wang University of Massachusetts, Amherst shufanwang@cs.umass.edu Mohit Iyyer University of Massachusetts, Amherst miyyer@cs.umass.edu Abstract Literary critics often attempt to uncover mean- ing in a single work of literature through care- ful reading and analysis. Applying natural lan- guage processing methods to aid in such lit- erary analyses remains a challenge in digital humanities. While most previous work fo- cuses on “distant reading” by algorithmically discovering high-level patterns from large col- lections of literary works, here we sharpen the focus of our methods to a single literary theory about Italo Calvino’s postmodern novel Invis- ible Cities, which consists of 55 short descrip- tions of imaginary cities. Calvino has provided a classification of these cities into eleven the- matic groups, but literary scholars disagree as to how trustworthy his categorization is. Due to the unique structure of this novel, we can computationally weigh in on this debate: we leverage pretrained contextualized representa- tions to embed each city’s description and use unsupervised methods to cluster these embed- dings. Additionally, we compare results of our computational approach to similarity judg- ments generated by human readers. Our work is a first step towards incorporating natural lan- guage processing into literary criticism. 1 Introduction Literary critics form interpretations of meaning in works of literature. Building computational models that can help form and test these interpre- tations is a fundamental goal of digital humani- ties research (Benzon and Hays, 1976). Within natural language processing, most previous work that engages with literature relies on “distant read- ing” (Jockers, 2013), which involves discover- ing high-level patterns from large collections of stories (Bamman et al., 2014; Chaturvedi et al., 2018). We depart from this trend by showing that computational techniques can also engage with lit- erary criticism at a closer distance: concretely, we Adelma (cities & the dead) … An old man was loading a basket of sea urchins on a cart; I thought I recognized him …he looked like a fisherman who, already old when I was a child, could no longer be among the living… Adelma is the city where you arrive dying and where each finds again… Eusapia (cities & the dead) … And to make the leap from life to death less abrupt, the inhabitants have constructed an identical copy of their city, underground… They say that every time they go below they find something changed in the lower Eusapia; the dead make innovations in their city; not many, but surely… Zobeide (cities & desire) …men of various nations had an identical dream. They saw a woman running at night through an unknown city… They dreamed of pursuing her… decided to build a city like the one in the dream… they settled, waiting for that scene to be repeated… Isidora (cities & memory) When a man rides a long time through wild regions he feels …

Content truncated to 500 words