Vector Semantics and Algorithmically-Assisted Close Reading
Abstract: The paper explores a point of convergence between literary theory and computational semantics. Using as its starting point the idea of deformance developed within the digital humanities, it seeks to harness a certain kind of reading machines to the purposes of literary criticism. These machines are conceptualized as cognitive models of some idealized readers and as such can provide insight into the requirements for the productive reading of specific kinds of texts. The theoretical underpinnings of the models are traced to the field of distributional semantics, and the computational – to that of natural language processing. The relevance of these vector space models is evaluated empirically via an algorithmically-assisted reading of a text from the genre of science fiction. The task is specifically selected so as to be a good testbed for the capability of the reading machines to actually model different kinds of readers, as predicted by literary theory.
Keywords: digital humanities, computational semantics, distributional semantics, natural language processing, science fiction studies
Alexander Popov, PhD, is a postdoctoral researcher in computational linguistics at the Bulgarian Academy of Sciences. He teaches a course on science fiction at the University of Sofia “St. Kliment Ohridski.” His research interests range over computational semantics, literary theory, modern philosophy, utopian studies and ecological criticism.
Alexander Popov – Vector Semantics and Algorithmically-Assisted Close Reading (pdf)