While work at the intersection of literary studies and GIS continues to grow in quantity and sophistication, there remains a frustratingly wide gulf between the promise of GIS’s analytical power and the difficulty of knocking literary datasets into the kind of shape which can be processed by GIS software. Point maps are manifestly inadequate for representing, much less analysing, the contingent, palimpsestic, ambiguous references by which places are habitually denoted in literary texts. Other modes of cartographic representation such as density smoothing solve some, though not all, of these problems. Various approaches have been proposed to address this problem of troublesome literary space: Gregory and Hardie’s ‘visual GISting’ (2011) combines GIS with collocate analysis, and more recently Gavin and Gidal (2017) have set out how a word-place matrix can be used to index a corpus to geographical features and thereby gain a sense of how discourses of ‘Ossianic space’ are geographically distributed across Scotland. In this paper, I use word embeddings on a corpus of 33 million words from the digitised Canadian periodical The Western Home Monthly to propose an approach which is able to better capture some of the complexities of the subtle interrelationships of places, terms, and spatialized concepts in literary texts. I also suggest some ways in which word embeddings can be used in conjunction with a GIS so as to represent some of the spatial and semantic interconnections between terms which are associated with spatial referents but which lack readily mappable co-ordinates.
References
Gavin, Michael, and Eric Gidal. “Scotland’s Poetics of Space: An Experiment in Geospatial Semantics.” Journal of Cultural Analytics, Nov. 2017
Gregory, Ian N., and Andrew Hardie. “Visual GISting: Bringing Together Corpus Linguistics and Geographical Information Systems.” Literary and Linguistic Computing, vol. 26, no. 3, Sept. 2011, pp. 297–314.