Edinburgh Research Explorer

Spatial dialectics: Pursuing geospatial imaginaries with word embedding models and mapping

Research output: Contribution to journalArticle

Related Edinburgh Organisations

Open Access permissions



Original languageEnglish
Issue number2
Publication statusPublished - 6 Jul 2019


As libraries and archives come under ever-greater pressure to digitize their holdings, especially collections such as periodicals where the fragility of the materials makes it preferable to offer users digitized versions rather than hard copies, it is increasingly incumbent upon scholars in the digital humanities to demonstrate the analytical gains to be made when performing computational analyses on digital cultural heritage data. The 260 digitized issues of the Western Home Monthly comprise one such dataset, presenting to the researcher both the tantalising possibility of discerning patterns of significance inaccessible by conventional close reading, and the sobering proposition that one cannot be sure ahead of time that the time spent preparing the data and correcting OCR errors will be worth the analytical payoff. In this paper, I use a series of digital humanities methods to approach the digitized Western Home Monthly to ask two related questions: first, the extent to which an archive with OCR problems can yield meaningful information when queried, and second, which digital approaches are tractable on periodical data. To do this, I focus on the geographic imaginary of the Western Home Monthly: its geographic points of reference and the spatiality of the imagined community it invokes. My argument is that by triangulating the results of different computational methods – corpus analysis, automated geoparsing, topic modelling and word vectors – it is possible to gain insights into the spatial imaginary of a periodical even where OCR errors and gaps in the data exist, and that these methods are of particular use in directing scholars to specific parts of a dataset where further analysis and OCR correction can be prioritised.

    Research areas

  • modernism, periodicals, digital humanities, word vectors, word embeddings

Download statistics

No data available

ID: 50434517