In a new paper lead by Mikael Brunila and accepted for publication and presentation at the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP) we introduce the concept of “critical toponymy” as an underexplored theoretical framework with valuable applications in applied natural language processing. This approach emphasizes recent advancements in embedding-based models that have improved the ability to identify place names in unstructured text data. While past work on toponymy has been largely approached as an engineering challenge, our work suggests that its application in addressing social scientific questions remains limited. We present preliminary steps in this direction by introducing novel data and a specialized Named Entity Recognition model to investigate the relationship between toponymic language, specifically in the context of the Airbnb market, and urban geospatial and sociocultural dynamics. Notably, we find that not all Airbnb hosts reference their units in relation to their residential neighborhoods, suggesting nuances in how urban dynamics are expressed through toponymic references. The study also identifies a correlation between toponymic references and gentrification dynamics, hinting at the possibility of a circular process where desirable toponyms attract guests, potentially accelerating gentrification processes driven by short-term rentals. While acknowledging the complexity of modeling such causal dynamics, the findings suggest that toponymic practices may play an active role in urban change beyond mere reflection.
The accepted pre-print is available here: https://arxiv.org/abs/2310.15302
Mikael Brunila, Jack LaViolette, Sky CH-Wang, Priyanka Verma, Clara Féré, Grant McKenzie. (2023) Toward a Critical Toponymy Framework for Named Entity Recognition: A Case Study of Airbnb in New York City. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP2023). December 6-10, 2023. Singapore.