Geospatiality of Topics in English Text Data

Kurzfassung

Many scientific disciplines have experienced an increased interest in the role of geographic location in what is often referred to as a “spatial turn”. Web texts that mention locations can be analyzed spatially, making them a promising data source. However, not all texts are equally likely to mention locations. The topic matters – but how much? Across several millions of texts from different web platforms and text types, we algorithmically (A) detected whether they contained an identifiable spatial location and (B) identified the general topic they contained, following a custom taxonomy of 19 topics derived in an inductive coding approach. We modeled the likelihood of geoinformation versus topic in a mixed-modeling approach, controlling for effects of author, time, and text length. A. Topic-specific effects on geospatiality were assessed via the topics’ fixed effects in the model. B. Similarity between geotagging and geoparsing approaches was analyzed using data from Twitter, where we compared the results based on geoparsed location mentions (non-geotagged) to results based on geotags. C. Similarity across datasets was measured using Spearman’s rank correlation coefficient. Results show: A. Observed frequency of mentioned locations varied strongly by topic, being high for the category Travel, Tourism & Migration and low for Private Life, Family & Relationships. The models attribute this largely to the effect of the topics. B. Substantially higher (and sometimes opposed) effects in geoparsingbased than in geotagging-based evaluation. C. Effects of topics largely correlated between datasets, moreso if datasets are similar in type and context. Conclusion: The topic of a text is a crucial factor in determining the presence of geolocation information. Recommendation: Consider and probe relationship between the analyzed content and frequency of georeferences to correctly interpret results. Outlook: Continue evaluation with different data, geoparsing methods, and different definitions of Geospatiality.