CoChat

Toponym-assisted map georeferencing: Evaluating the use of toponyms for the digitization of map collections

Open access
Bahgat, Karim, Runfola, Dan · 2021 · PLOS ONE

The DOI matches its Crossref record; 84 of the 85 references it cites were found in a registry or an index; none of the 45 sources its claims cite is checked for support yet.

Abstract

A great deal of information is contained within archival maps—ranging from historic political boundaries, to mineral resources, to the locations of cultural landmarks. There are many ongoing efforts to preserve and digitize historic maps so that the information contained within them can be stored and analyzed efficiently. A major barrier to such map digitizing efforts is that the geographic location of each map is typically unknown and must be determined through an often slow and manual process known as georeferencing. To mitigate the time costs associated with the georeferencing process, this paper introduces a fully automated method based on map toponym (place name) labels. It is the first study to demonstrate these methods across a wide range of both simulated and real-world maps. We find that toponym-based georeferencing is sufficiently accurate to be used for data extraction purposes in nearly half of all cases. We make our implementation available to the wider research community through fully open-source replication code, as well as an online georeferencing tool, and highlight areas of improvement for future research. It is hoped that the practical implications of this research will allow for larger and more efficient processing and digitizing of map information for researchers, institutions, and the general public.

Overview

CoChat Summary

TL;DR

This paper presents and evaluates a fully automated toponym-based method for georeferencing maps by extracting place-name labels, matching them to gazetteers, and estimating a transform. Across simulated and real-world maps, it finds the approach can georeference many maps accurately enough for data extraction, though success depends on toponym count and image resolution.

Key findings

  • Automated toponym-based georeferencing achieved high accuracy (<5% normalized maximum error) for 26.9% of 6,430 simulated maps and low accuracy (<100% error) for 77.7%; for a realistic simulated subset (n = 4,390), 36.7% were high accuracy and 86.7% low accuracy, and for a high-resolution realistic subset (n = 3,512), 40.1% were high accuracy and 91.4% low accuracy.
  • On 333 real-world country maps, the modelMaxLOO metric indicated 67% within <5% error, 44% within <1% error, 82% within <100% error, and 15.3% failed.
  • Fewer toponyms and coarse resolution most affected low-accuracy success: maps with only 10 toponyms dropped from about 80% success to 40%, and 1000-pixel resolution dropped success from over 80% to 60%. Toponym uncertainty up to 10 km and map extents as small as 25 km remained workable.
  • Model selection favored lower-order polynomials: 1st order was used in 69% of georeferenced maps, 2nd order in 24%, and 3rd order in 7%.

Methods

  • The pipeline extracts toponym labels using CIELab Delta E 2000 color thresholding (ΔE < 25 from black), Tesseract sparse text recognition, rules-based grouping of words within approximately 1.5 font-height, and contour-based detection of toponym marker symbols; toponyms are geocoded against USGS GNS, GeoNames, CIESIN GRUMP, OSMNames, and Natural Earth gazetteers.
  • Toponym disambiguation uses normalized point pattern matching and combinatorial optimization over triplets, expanding candidate result sets; transform estimation compares 1st, 2nd, and 3rd order polynomial transforms and selects among control-point sets using leave-one-out maximum residual (modelMaxLOO) with automated outlier removal.
  • Evaluation used 7,580 simulated maps generated from 379 scenes with varied extent, projection, toponym count, toponym uncertainty, resolution, and pixel noise, plus 333 real-world country maps from the University of Texas at Austin Map Collection (average width approximately 1300 pixels); simulated accuracy used trueMax, while real-world accuracy used modelMaxLOO.

Limitations

  • The paper states the algorithm is not guaranteed to find all matching sets, unlike an exhaustive search, and may produce spurious matches; it cautions against use for maps smaller than 25 km due to sparse meaningful toponyms, and notes projection testing was limited to common projections, so results should not be generalized to less common source projections.
  • The paper states that text recognition, geocoding/matching, and transform selection remain areas for improvement; false matched toponyms and distorted transformations occurred, and real-world maps lack true coordinates, so accuracy was measured by modelMaxLOO rather than trueMax.
  • Evident from the text: the full simulated sample excluded edge-case parameter combinations, and 1000-pixel jpg maps were dropped because labels were illegible; the real-world sample was limited to country maps from one collection at low-to-medium resolution, so generalization to other map types or collections is not established.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1371/journal.pone.0260039

Cites

84 of 84 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig 1 Example of a semi-automated workflow. Screenshot of a toponym-assisted georeferencing workflow. In a semi-automated workflow, if a user notices problems with toponym-based georeferencing, adjustments can be made through the manual selection and correction of control points. The map image is from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and is in the public domain.
Fig 1 Example of a semi-automated workflow. Screenshot of a toponym-assisted georeferencing workflow. In a semi-automated workflow, if a user notices problems with toponym-based georeferencing, adjustments can be made through the manual selection and correction of control points. The map image is from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and is in the public domain.CC BY · PubMed Central
Fig 2 Illustration of toponym recognition. (a) Thresholded image using the △E metric for the automated identification of text and toponym markers. (b) Identified map text labels (in green) and toponym marker points (in red). The map image is from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and is in the public domain: https://maps.lib.utexas.edu/maps/middle_east_and_asia/china_pol96.jpg.
Fig 2 Illustration of toponym recognition. (a) Thresholded image using the △E metric for the automated identification of text and toponym markers. (b) Identified map text labels (in green) and toponym marker points (in red). The map image is from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and is in the public domain: https://maps.lib.utexas.edu/maps/middle_east_and_asia/china_pol96.jpg.CC BY · PubMed Central
Fig 3 Pattern-based toponym identification. (a) Point pattern of the toponyms Mbe, Poli, and Tchollire detected in the image. (b) Point pattern of the corresponding geographic coordinates to the toponyms. (c) Full view of all candidate gazetteer matches and their point patterns. The map image is from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and is in the public domain: https://maps.lib.utexas.edu/maps/africa/cameroon_pol98.jpg. The geodata used to render country outlines is from ©Natural Earth data and is in the public domain.
Fig 3 Pattern-based toponym identification. (a) Point pattern of the toponyms Mbe, Poli, and Tchollire detected in the image. (b) Point pattern of the corresponding geographic coordinates to the toponyms. (c) Full view of all candidate gazetteer matches and their point patterns. The map image is from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and is in the public domain: https://maps.lib.utexas.edu/maps/africa/cameroon_pol98.jpg. The geodata used to render country outlines is from ©Natural Earth data and is in the public domain.CC BY · PubMed Central
Fig 4 Normalized point pattern coordinates. Pattern matching of image placename toponyms Mbe, Poli, and Tchollire and coordinate combinations in normalized space.
Fig 4 Normalized point pattern coordinates. Pattern matching of image placename toponyms Mbe, Poli, and Tchollire and coordinate combinations in normalized space.CC BY · PubMed Central
Fig 5 Bounding boxes of the simulated map scenes. The geodata used to render country outlines is from ©Natural Earth data and is in the public domain.
Fig 5 Bounding boxes of the simulated map scenes. The geodata used to render country outlines is from ©Natural Earth data and is in the public domain.CC BY · PubMed Central
Fig 6 Simulated map accuracy categories. Shows example georeferenced maps overlaid on source maps with known coordinates. Map accuracy is calculated based on normalized maximum map error (as a percentage of image radius). The simulated maps were generated based on public domain data from ©Natural Earth, including data on country outlines, populated places, urban extents, rivers, and roads.
Fig 6 Simulated map accuracy categories. Shows example georeferenced maps overlaid on source maps with known coordinates. Map accuracy is calculated based on normalized maximum map error (as a percentage of image radius). The simulated maps were generated based on public domain data from ©Natural Earth, including data on country outlines, populated places, urban extents, rivers, and roads.CC BY · PubMed Central
Fig 7 Effect of simulated map parameters for georeferencing success rates. Share of all simulated maps that were successfully georeferenced, for different parameter values.
Fig 7 Effect of simulated map parameters for georeferencing success rates. Share of all simulated maps that were successfully georeferenced, for different parameter values.CC BY · PubMed Central
Fig 8 Georeferencing results for real-world country maps. Shows subset of maps with errors less than 5% of map radius, representing about 67% of the total sample. The georeferenced overlay maps are from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and are in the public domain. The complete list of all map images and their source URLs can be found in the replication data accompanying this article (see Data Availability statement). The geodata used to render country outlines is from ©Natural Earth data and is in the public domain.
Fig 8 Georeferencing results for real-world country maps. Shows subset of maps with errors less than 5% of map radius, representing about 67% of the total sample. The georeferenced overlay maps are from the University of Texas at Austin’s Perry-Castañeda Library (PCL) Map Collection and are in the public domain. The complete list of all map images and their source URLs can be found in the replication data accompanying this article (see Data Availability statement). The geodata used to render country outlines is from ©Natural Earth data and is in the public domain.CC BY · PubMed Central

Verification

Claims the paper makes

0 Valid · 0 Not valid · 17 Not checked · 60 Own results and interpretations

39 more claims

Create a free account to see all 77 claims and to verify the 39 references still open.

See all claims — free account
  • Probably not a substantial claim. 18
  • It may state two separate claims as one. 22

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 84 references

Research works printed but not linked: 8

  1. MJ Jackson, 1991, Geographical Information Systems: Principles and Applications—Volume 1: Principles, 239
  2. P Bolstad, 2014, GIS Fundamentals: a first text on Geographic Information Systems
  3. G Kurt Menke, 2016, Mastering QGIS
  4. MJ Goodchild, 1991, Geographical Information Systems: Principles and Applications—Volume 1: Principles, 40
  5. Herold H, Roehm P, Hecht R, Meinel G. Automatically georeferenced maps as a source for high resolution urban growth analyses. In: Proceedings of the ICA 25th International Cartographic Conference. 1; 2011. p. 1–5.
  6. M Rusiñol, 2011, Automatic index generation of digitized map series by coordinate extraction and interpretation, e-Perimetron, 6, 219
  7. Leyk S, Chiang YY. Information extraction based on the concept of geographic context. In: Proc. AutoCarto; 2016. p. 100–110.
  8. Inoue R, Wako M, Shimizu E. A new map transformation method for highly deformed maps by creating homeomorphic triangulated irregular network. In: XXIII International Cartographic Conference. Moscow, Russia. DVD; 2007.

Cited by

See all 16

Metadata

Authors
Bahgat, K., & Runfola, D.
Affiliations
William & Mary (US)
Venue
PLOS ONE
Published
2021-11-18
Volume
16
Issue
11
Pages
e0260039
ISSN
1932-6203 (ISSN-L)
OpenAlex
W3212834916
Cited by
16
References
85
Citations by year
2026: 9 · 2025: 4 · 2024: 1 · 2023: 1 · 2022: 1
FWCI
1.24
Type
journal article
Language
en
Open access
Locations
  • PLoS ONE (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • W&M Publish (College of William & Mary) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Bill and Melinda Gates Foundation
Topics
  • Social Sciences › Geography, Planning and Development › Historical Geography and Cartography
  • Social Sciences › Arts and Humanities › Literature and Literary Theory › Digital Humanities and Scholarship
  • Social Sciences › Geography, Planning and Development › Geographic Information Systems Studies
Keywords
place names, map digitization, archival maps, historical cartography
MeSH terms
Data Collection, Humans, Image Processing, Computer-Assisted, Maps as Topic, Pilot Projects, Software, Information Storage and Retrieval, Geographic Mapping

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-19, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-19 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Toponym-assisted map georeferencing: Evaluating the use of toponyms for the digitization of map collections. https://app.cochat.ai/papers/10.1371/journal.pone.0260039
The paper
Bahgat, K., & Runfola, D. (2021). Toponym-assisted map georeferencing: Evaluating the use of toponyms for the digitization of map collections. PLOS ONE. https://doi.org/10.1371/journal.pone.0260039

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.