Deep Learning for Toponym Resolution: Geocoding Based on Pairs of Toponyms

Abstract

Geocoding aims to assign unambiguous locations (i.e., geographic coordinates) to place names (i.e., toponyms) referenced within documents (e.g., within spreadsheet tables or textual paragraphs). This task comes with multiple challenges, such as dealing with referent ambiguity (multiple places with a same name) or reference database completeness. In this work, we propose a geocoding approach based on modeling pairs of toponyms, which returns latitude-longitude coordinates. One of the input toponyms will be geocoded, and the second one is used as context to reduce ambiguities. The proposed approach is based on a deep neural network that uses Long Short-Term Memory (LSTM) units to produce representations from sequences of character n-grams. To train our model, we use toponym co-occurrences collected from different contexts, namely textual (i.e., co-occurrences of toponyms in Wikipedia articles) and geographical (i.e., inclusion and proximity of places based on Geonames data). Experiments based on multiple geographical areas of interest—France, United States, Great-Britain, Nigeria, Argentina and Japan—were conducted. Results show that models trained with co-occurrence data obtained a higher geocoding accuracy, and that proximity relations in combination with co-occurrences can help to obtain a slightly higher accuracy in geographical areas with fewer places in the data sources.

Toward a Critical Toponymy Framework for Named Entity Recognition: A Case Study of Airbnb in New York City

“…unstructured text data. Thus far, such work in the context of toponymy has generally been explored as an engineering challenge to be improved upon with increasingly sophisticated methods (e.g., Cardoso et al, 2022;Davari et al, 2020;Fize et al, 2021;Tao et al, 2022). While such efforts are invaluable to engineers and social scientists alike, few have extended these novel approaches to concrete questions of social scientific intrigue.…”

Multi‐unit building address geocoding: An approach without indoor location reference data

“…Recent work focuses on the improvement of the reference datasets required by the latter two methods. The work uses state‐of‐the‐art machine learning models to either optimize address naming to improve address point mapping (Fize et al, 2021; Lee et al, 2020; Matci & Avdan, 2018) or learn from images to more accurately identify parcels and other shapes for polygon mapping (Laumer et al, 2020; Wang et al, 2018; Yin et al, 2019). No work that suggests another method to geocode addresses in the absence of such detailed reference data has been found and neither method can accurately geocode addresses located in multi‐unit buildings.…”

How can voting mechanisms improve the robustness and generalizability of toponym disambiguation?

“…Context features refer to the contextual similarity between the toponym and the candidate, while the context of the candidate is obtained from its Wikipage. Apart from fully supervised approaches, weakly-supervised and unsupervised approaches (Speriosu and Baldridge 2013;Ardanuy and Sporleder 2017;Fize, Moncla, and Martins 2021) have also been proposed to reduce the amount of annotated data required. For example, Ardanuy and Sporleder (2017) define a model to score and rank candidates, using features like the context similarity between a toponym and a candidate, geographic closeness to the base location of a collection, and geographic closeness of toponyms.…”