Which Melbourne? Augmenting Geocoding with Maps

Abstract

The purpose of text geolocation is to associate geographic information contained in a document with a set (or sets) of coordinates, either implicitly by using linguistic features and/or explicitly by using geographic metadata combined with heuristics. We introduce a geocoder (location mention disambiguator) that achieves state-of-the-art (SOTA) results on three diverse datasets by exploiting the implicit lexical clues. Moreover, we propose a new method for systematic encoding of geographic metadata to generate two distinct views of the same text. To that end, we introduce the Map Vector (MapVec), a sparse representation obtained by plotting prior geographic probabilities, derived from population figures, on a World Map. We then integrate the implicit (language) and explicit (map) features to significantly improve a range of metrics. We also introduce an open-source dataset for geoparsing of news events covering global disease outbreaks and epidemics to help future evaluation in geoparsing.

Text Meets Space: Geographic Content Extraction, Resolution and Information Retrieval

“…In this tutorial, we aim to give a survey of the concepts and methods used to make implicit spatial evidence contained in text collections accessible. We cover selected early and seminal attempts [3,8,10,13] and more recent Machine Learning (ML) methods [6,[16][17][18], hoping to inspire students and fellow researchers to get interested in conducting their own research in this area. Bringing two seemingly disparate worlds like geographic space and text documents together is exciting!…”

Enhancing spatial and textual analysis with EUPEG: An extensible and unified platform for evaluating geoparsers

“…A simple method is to resolve a place name to the place instance that has the highest population or the largest total geographic area (Ladra, Luaces, Pedreira, & Seco, ; Li, Srihari, Niu, & Li, ). Machine learning models have also been developed for toponym resolution by exploiting various features, such as toponym co‐occurrences (Overell & Rüger, ), words in the local context (Speriosu & Baldridge, ), distances among toponyms (Santos, Anastácio, & Martins, ), topics of the local context (Ju et al., ), and a combination of multiple features (Gritta et al., ; Nesi, Pantaleo, & Tenti, ).…”

“…developed by Gritta et al. (), which integrates convolutional neural networks, word embeddings, and the geographic vector representations of place names. There also exist commercial geoparsers, such as Geoparser.io (https://geoparser.io), which often charge a fee.…”

“…Gritta, Pilehvar, and Collier () and Gritta et al. () shared two human annotated corpora, GeoVirus and GeoWebNews, which contain 229 and 200 news articles, respectively. TR‐News is another news article corpus which contains 118 articles manually annotated by Kamalloo and Rafiei ().…”

“…While a number of geoparsers have already been developed, it is difficult to directly compare their performances. Examples of geoparsers developed include MetaCarta (Frank, Rauch, & Donoghue, ), GeoTxt (Karimzadeh et al., ), the Edinburgh Geoparser (Alex, Byrne, Grover, & Tobin, ), TopoCluster (DeLozier, Baldridge, & London, ), and CamCoder (Gritta, Pilehvar, & Collier, ). Two factors make direct comparison difficult.…”