Rediscovery as validation: unsupervised analysis of Iberian place names recovers known dialect areas, isoglosses, and historical frontiers

Preprint (author's original version), not peer reviewed. Submitted to Digital Scholarship in the Humanities (Oxford University Press) on 15 November 2026. Purpose. Unsupervised methods find spatial patterns in any large gazetteer, so they must show that their patterns are real. We test 'rediscovery as validation': before discovering anything, a method should recover, unprompted, what is already known. Design/methodology/approach. About 840,000 Iberian place names from OpenStreetMap, GeoNames, and the official gazetteer of Spain form four corpora. Word beginnings and endings are extracted without any list of suffixes, factorized into spatial layers by non-negative matrix factorization, scanned for local excesses, and scored against twenty-one reference dialect areas with explicit chance levels. Findings. Selecting elements by effect size rather than by significance raises the mean number of areas recovered from 5.7 (chance 1.4) to 8.7 (chance 3.1), and nine areas replicate in at least two corpora. The scan places 90% of the elements of known strata in their expected regions (random elements: 18%). Targeted measurements reproduce the Castilian wedge as two isoglosses, a two-step northern edge of Arabic toponymy, and the concentration of -ena in Baetica, which a Basque homograph hides in the official gazetteer. Originality. To our knowledge, this is the first study to combine unsupervised element extraction, spatial factorization, and scan statistics over a whole peninsula, and to score the output against known areas with chance levels. Contribution to the field of Digital Humanities. We offer a reproducible protocol for validating exploratory methods on unevenly documented cultural data, show that significance-based selection favours densely mapped regions, and rank new leads, such as the -aque/-eque names around Toledo, for documentary study. Code and data: https://github.com/carlos-vi/iberian-toponymy-rediscovery

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-15
DOI
https://doi.org/10.5281/zenodo.24246163
Primary Topic
Spanish Linguistics and Language Studies
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Rediscovery as validation: unsupervised analysis of Iberian place names recovers known dialect areas, isoglosses, and historical frontiers

Carlos Vidal-Perez
Zenodo (CERN European Organization for Nuclear Research)
Spanish Linguistics and Language Studies
preprint

Rediscovery as validation: unsupervised analysis of Iberian place names recovers known dialect areas, isoglosses, and historical frontiers

Carlos Vidal-Perez
preprint en

Abstract

Preprint (author's original version), not peer reviewed. Submitted to Digital Scholarship in the Humanities (Oxford University Press) on 15 November 2026. Purpose. Unsupervised methods find spatial patterns in any large gazetteer, so they must show that their patterns are real. We test 'rediscovery as validation': before discovering anything, a method should recover, unprompted, what is already known. Design/methodology/approach. About 840,000 Iberian place names from OpenStreetMap, GeoNames, and the official gazetteer of Spain form four corpora. Word beginnings and endings are extracted without any list of suffixes, factorized into spatial layers by non-negative matrix factorization, scanned for local excesses, and scored against twenty-one reference dialect areas with explicit chance levels. Findings. Selecting elements by effect size rather than by significance raises the mean number of areas recovered from 5.7 (chance 1.4) to 8.7 (chance 3.1), and nine areas replicate in at least two corpora. The scan places 90% of the elements of known strata in their expected regions (random elements: 18%). Targeted measurements reproduce the Castilian wedge as two isoglosses, a two-step northern edge of Arabic toponymy, and the concentration of -ena in Baetica, which a Basque homograph hides in the official gazetteer. Originality. To our knowledge, this is the first study to combine unsupervised element extraction, spatial factorization, and scan statistics over a whole peninsula, and to score the output against known areas with chance levels. Contribution to the field of Digital Humanities. We offer a reproducible protocol for validating exploratory methods on unevenly documented cultural data, show that significance-based selection favours densely mapped regions, and rank new leads, such as the -aque/-eque names around Toledo, for documentary study. Code and data: https://github.com/carlos-vi/iberian-toponymy-rediscovery

Zenodo (CERN European Organization for Nuclear Research)
Spanish Linguistics and Language Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
Rediscovery as validation: unsupervised analysis of Iberian place names recovers known dialect areas, isoglosses, and historical frontiers — Carlos Vidal-Perez · Zenodo (CERN European Organization for Nuclear Research) (2026) | TGRS Research Map | TGRS