Disambiguating spatial prepositions: The case of geo‐spatial sense detection

Abstract

Spatial relations in natural language are frequently expressed through prepositions. Thus, in the locative expressions “New York in the United States” and “the house on the river” the prepositions “in” and “on,” respectively, serve to communicate the relationships in space between the subject and object of the preposition. Automatic detection of the use of prepositions in a spatial and in particular a geo‐spatial sense that refers to geographic context is of interest in supporting automated methods for determining the actual geographic location referred to by locative expressions. This work focuses on disambiguation of prepositions in natural language, with the goal of distinguishing whether a preposition is used in a specifically geo‐spatial sense. We conduct machine learning experiments that demonstrate the clear benefit for geo‐spatial sense detection of using transformer model deep learning methods when compared with a variety of methods, that include Naive Bayes, support vector machine, and random forest classifiers with handcrafted linguistic features, and a bag of words approach with a meta‐classifier that adds geo‐spatial features. The best performance was obtained with the Bidirectional Encoder Representation from Transformer‐based XLNet transformer model, with a best precision of 0.96 and an F1 score of 0.94 when evaluated on a corpus of natural language expressions that were annotated for this task. We also conducted experiments to detect generic spatial sense, in which the best F1 score, of 0.95, was again obtained with XLNet.

Geo-knowledge-guided GPT models improve the extraction of location descriptions from disaster-related social media messages

“…There exists another thread of related research on detecting geospatial or, more generally, spatial descriptions from natural language text (Liu, Vasardani, and Baldwin 2014;Stock, Jones, and Tenbrink 2022). While often studying spatial descriptions under more general contexts (e.g., daily life), research in this thread, such as disambiguating geospatial prepositions (Radke et al 2022), can be very useful for the step of geo-locating the recognized location descriptions.…”

“…Also, both models extract whole door number addresses including the prepositions (e.g., "in") sometimes used by people but not typically part of formal addresses. These extracted whole addresses allow us to analyze and geo-locate them in the next step by building on previous research, such as research on geospatial preposition analysis (Radke et al 2022). Given that door number addresses provide precise location information, correctly recognizing location descriptions in this category can be highly helpful in identifying the locations of victims.…”

The BioWhere Project: Unlocking the Potential of Biological Collections Data

“…We are comparing a range of NER methods, including spaCy and several BERT-based transformer deep learning models (BERT, DistilBERT, ALBERT, RoBERTa, XLNet), which we are retraining using our own geographic corpora. In this we are following earlier work (Radke et al, 2022) that demonstrates that these methods outperform previous approaches to the detection of spatial relation terms (Kordjamshidi et al, 2011;Radke et al, 2019). Following extraction of the entities, we are developing methods for relation extraction to incorporate relevant geographic context.…”