Improving Toponym Resolution with Better Candidate Generation and Transformer-based Reranking

Anonymous

Abstract: Geocoding is the task of converting location mentions in text into structured geospatial data.We propose a new architecture for geocoding, SSPART, that first uses information retrieval techniques to generate a list of candidate entries from the geospatial ontology, and then reranks the candidates using a transformer-based neural network.The reranker compares the location mention to each candidate entry, while incorporating additional information such as the entry's population, the entry's type of location, and the sentences surrounding the mention in the text.Our proposed toponym resolution framework achieves state-of-the-art performance on multiple datasets.Code and models are available at \url{https://<anonymized>}.

Paper Type: long

Research Area: Information Extraction