=Paper=
{{Paper
|id=Vol-2364/24_paper
|storemode=property
|title=Historical Settlement Units as Linked Open Data
|pdfUrl=https://ceur-ws.org/Vol-2364/24_paper.pdf
|volume=Vol-2364
|authors=Olof Karsvall
|dblpUrl=https://dblp.org/rec/conf/dhn/Karsvall19
}}
==Historical Settlement Units as Linked Open Data==
https://ceur-ws.org/Vol-2364/24_paper.pdf
Historical Settlement Units as Linked Open Data
Olof Karsvall1[0000-0003-4350-7533]
1
The Swedish National Archives • TORA
Abstract. TORA is a historical geographic register provided by the Swedish Na-
tional Archives. It is built around the concept Historical settlement unit, referring
to villages, hamlets and farms in the medieval and early modern rural society.
Currently, approximately fifty percent of all such units in Sweden are covered,
which are mentioned in written sources and maps before c. 1800. This article
discusses and presents the definition and the method of the core concept. Using
linked data technologies, the goal is to increase, simplify and improve the use of
historical sources in research and applications.
Keywords: historical settlement unit, linked open data, gazetteer.
1 Introduction
Most data on the Web can be related to places. Place as a resource on the Semantic Web
is therefore a fundamental component, which sets data about people, events and other
things into a geographical context. Location data make delimitation of large data sets
possible. It also enables spatial presentations and analyzes, and thus is a complement
to lists and graphs on the linked data cloud.
The availability of spatial data today is however limited. Authorities and other actors
offers free map services but rarely exposes the underlying data, which is necessary for
places to be linkable resources. Map services or gazetteer, providing place names and
geolocations, such as GeoNames and Open Street Maps, partially fills this gap, alt-
hough data quality is not guaranteed, as these services are aggregations of various data
sources of different kind, and furthermore require CC-BY licenses or similar for re-use.
The lack of geographical data applies not least to historical places. Few map services
handle historical locations, which means they are not particularly useful in relation to
historical sources and research. Of course, many historical places — not least towns
and major settlements — are still places today. In situations where the aim is to get an
overview of historical data, it might not be necessary to have accurate historical refer-
ences and positions. But in other cases, when a place referred in a historical document
is to be defined and described, then proper historical geo data are required.
260
To allow accurate historical geocoding of places, exposed as linked data nodes on
the Semantic Web, the concept Historical Settlement Unit has been created. The initia-
tive has been taken within the research project TORA at the Swedish National Ar-
chives.1
The goal is to establish a comprehensive register that could be used by various re-
searchers as an authority register or gazetteer, which reduces the need for others to
create their own historical place registers. A further purpose is to encourage geograph-
ical and historical surveys. It’s exposed in the linked data format (RDF) and could be
used in various search application and, for example, with named entity recognition
tools.2 At present the register covers 27 000 settlements, which corresponds to between
one third and half of all settlements during the Middle Ages until the period around c.
1800 (see Fig. 2).
This article has a methodological aim: first to define the core concept, thereafter to
describe the data model and delimitations. Similarities and differences with other his-
torical gazetteer will be discussed throughout. The main question raised is: How could
a historical-geographical register and concept be created, that meets the need of being
intelligible on the Semantic Web, while fulfilling scientific criteria, as source criticism
and accuracy, that makes it relevant as a resource in research?
2 The definition of Historical Settlement Unit
From the Middle Ages until the agriculture and industrial revolution of the 18th and
19th century, the society consisted exclusively of rural villages, hamlets and farms.
Historians have emphasized that the village organization, as we know it from historical
sources, began to be established during the 9th and 10th centuries; a system combining
individual ownership of land with collaboration farming practices.3 With an increasing
amount of sources from the High Middle Ages, as e.g. law books and cadasters, these
settlements appear as spatial or territorial units, with rules for how land were distributed
within the villages and with defined boundaries towards neighboring villages.
The purposed definition of Historical settlement unit is based on the medieval and
early modern rural conditions where settlements, such as villages, hamlets and single
farms, appeared as basic spatial, functional and economic units. Also towns and larger
farms, such as demesnes belonging to lords, churches and monasteries, could constitute
such a unit. Although they differ in size and function they appear in a similar way as
unit in the historical rural landscape. Also abandoned villages and farms, that often are
vaguely documented but likely had been settled, could be classified as Historical set-
tlement units.4
1
Since 2003, a major work has been carried out at the National Archives in Stockholm con-
cerning digitalization of the oldest large-scale maps from the 17th century over Sweden
(https://riksarkivet.se/geometriska).
2
An example of this is discussed and described in Karsvall & Borin 2018.
3
See e.g. Wickham 2005 p 514–518.
4
Karsvall 2011 p 22; 2016.
261
A demarcation around year 1800 is needed since, as a result of a major landscape
changes, the historical settlement units are less relevant. Land reforms, industrialization
and urbanization did re-shape and re-define many settlements. Not least the enclosure
land reform Laga skifte in Sweden, which resulted in a new division of the land, had
great impact on settlement patterns, as villages and hamlets more or less ceased being
functional units and instead became groups of individual farms. Although the place
names often were kept, the spatial definition of a historical place is not necessarily rel-
evant after the last two centuries rapid changes. In other words, the slow development
in landscapes and settlements – the long duration or La longue durée to use the term of
Fernand Braudel (1958) – motivates a creation of ‘historical’ settlement as a concept
for the Semantic Web.
3 Creation of gazetteer
A historical gazetteer can be created in different ways, either from the ground up on the
basis of original sources such as maps and written sources or fully or partly based on
published secondary sources. The conditions for establishing a register vary between
different countries depending on access to source material, especially historical maps
and related geographic and topographical sources.
Researchers can use historical atlases and other reference works about places.
Digitization of such books would simplify access but not improve usage.5 A greater
benefit is achieved when the reference works go through a digitalization process and
becomes machine-readable using a data model and place identities, that various sources
could refer to. The development of geographical applications (GIS) has considerably
simplified the ability to create, analyze and publish geographic data. Different methods
have been examine in related to geocoding of historical sources. Research projects have
put focus in creating digital historical atlas (“Historical GIS”, or “HGIS”) developing
Web-GIS search applications.6
In recent years, initiatives have been taken to establish digital geographical registers
or gazetteers, and services that use them.7 Open data formats and aggregation of geo-
data has gained greater attention. Also crowd-sourcing has been used as a method of
transcribing and converting large amounts of text sources into data.8
5
E.g. https://sok.riksarkivet.se/rosenberg.
6
Examples of such projects: National Edition of the Oldest Geometrical Maps 1630–1655,
https://riksarkivet.se/geometriska; SVEA-Pommern Karten und Texte der Schwedischen
Landesaufnahme von Pommern 1692–1709, http://www.svea-pommern.de; Digitalt atlas over
Danmarks historisk-administrative geografi, http://digdag.dk, see also http://www.hgis.org.uk
for a list of other similar resources.
7
E.g. http://commons.pelagios.org; http://whgazetteer.org.
8
The GB1900 Gazetteer, which publishing place names from the period around 1900, is likely
one of the largest datasets of this kind so far online, holding over 1.7 million identifications
and related coordinates, created through crowd-sourcing (Aucott et al 2018; Southall et al
2017; https://geo.nls.uk/maps/gb1900.
262
The editorial work of creating a register or gazetteer of Historical settlement units
have so far (in the TORA project) considered the conditions in the Nordic countries and
especially Sweden. There are two comprehensive source materials that enables accurate
or precise definitions of historical places in Sweden: firstly, the Crown’s cadastres from
the 1530s and beyond; secondly the land survey maps from the 1630s and onwards.9
When used together, these two sources are invaluable when creating a historical gazet-
teer. The cadastre reports almost all villages and farms by name in a uniform way for
the whole Sweden by mid 16th century. The survey maps, especially the oldest series
from 17th and 18th century, provides a detailed view of the rural landscape. Without
access to historical maps, and primary large-scale maps, there is hard to make precise
identification of historical places. Although, as an alternative, it could be possible to
reconstruct the locations of settlements by combining data from older written sources
with more present maps.10 It is nevertheless the great availability of historical maps in
Sweden that is the foundation of Historical settlement unit, that is described in more
detail below.
4 Point objects vs. polygons: How the spatial coordinates are set
So far we have specified what kind of object Historical settlements units are and their
time period. In the following, a deeper insight is given on how the geographical coor-
dinates are extracted, that allow these historical records to be exposed as spatial data
objects.
As mentioned, settlement such as villages and hamlets were legally defined by prop-
erty boundaries, which separated one unit from other surrounding units. One way to
define these would be to reconstruct the external property boundary. In practice, how-
ever, this is rarely possible. Two reasons can be stated. Firstly, it is difficult and not
always possible to interpret and transfer the boundaries on a historical map into a mod-
ern map. It assumes that all boundaries are visible on older maps, which seldom is the
case. Secondly, it takes far too much time to make complete rectifications of village
boundaries.
In a gazetteer, it is reasonable to define administrative divisions, such as parishes,
using polygon boundaries, but not rural settlements. Boundary rectification could be an
aim of a research projects, but not the foundation in a digital register holding thousands
of settlements. Consequently, it is better to define historical settlement as point objects,
i.e. as coordinate pairs (latitude and longitude).
The choice of using point objects requires a method; a definite way for how the
coordinate pair are set. A settlements such as a village forms an area of one or more
square kilometers. There are therefore several alternatives locations where the coordi-
nate points could be set. The official map service of place names in Sweden today holds
coordinates of place name positions. The aim is cartographic, and the place names have
9
For further information on these sources, see, for example Kain & Baigent 1992 p 49; Karsvall
2007; 2016.
10
See e.g. Panecki et al 2018.
263
been set next to the settlement area to improve visibility. Such principle is less suitable
for defining actual settlement locations.11
A better way to define historical settlements would be to define them as spatial (ge-
ographical) units, focusing on the cultivate land, and primary the most valuable land,
as in most cases the arable land. Only one coordinate pair is to be set for each settlement
unit. Since there could be several farmsteads within a settlement is it not suitable to use
the location of farmsteads. Individual farms could be disbanded or displaced, although
the settlement as a unit remains. The farmsteads itself could also carry their own coor-
dinates in other registers, and should not to be mixed with the settlement unit. One way
would be to put the coordinate as centrally as possible. The problem with such a prin-
ciple is, however, that both farms and land often appear at a certain part of the village,
which rarely represents the center of the settlement area.
A better way to define settlements is therefore to use the historical map available,
and set the coordinates at a location that according to the map constitutes the core cul-
tivated part of the farmland. From a geographical and economic perspective, this loca-
tion correlates to the most valuable land, usually the arable land lying next to the build-
ing plots. This means that surrounding land, such as forests and pastures, will be irrel-
evant for the spatial definition, which instead aims to point out the central part of the
settlement. In situations when farmsteads are dispersed or when there are several scat-
tered arable fields, the coordinates are set at the assumed core arable area.
For identification of the settlements, land survey maps created before c. 1800 are
used in the first place. In the absence of such older historical maps, identifications can
also be done (but with less accuracy) with the help of more modern maps from 19th
century and later. Such a procedure – using much younger maps – which could be per-
ceived anachronistic, does not have to be so. By adding the criterion the place should
be mentioned in a historical written source (before c 1800), there are reason to believe
the settlement found the later map (holding the same place name) is the same object.
Rural settlements have often, at least in the Nordic countries, been stable over extensive
period of time. Most changes occurred within the villages – the number of farms could
increase or decreased during periods of expansion and crises, but the external bounda-
ries that form the settlement as a unit were often kept intact. Even during periods of
decline, settlements have often been re-settled.12
To conclude this discussion: The aim of setting coordinates to Historical settlement
units is to identify what likely was the accurate location during the middle ages and/or
the early modern period. The use of later modern maps for identification results in
greater uncertainty. Two statements should be emphasized. Firstly, one coordinate pair
(latitude and longitude) represent a settlement unit, more precisely the farm land –
roughly one or a few square kilometers. Secondly, the coordinates are set at a central
location, relative to the core cultivated land, nearby the farmsteads and nearby arable
land.
11
The official place names in Sweden is searchable at https://kso.etjanster.lantmateriet.se.
12
The method of analyzing older historical circumstances on basis of younger sources is well
proven and known as retrogressive by historical geographers (see e.g. Karsvall 2013 p 411).
264
5 Data model and vocabulary
The registration of historical settlement units in Sweden started in 2003 in a previous
project, but the data model and the format have recently been adjusted to linked data
standards, the Resource Description Framework (RDF). This section will describe the
adaptation into linked data standards and the design choices made. There are two main
reasons why RDF has been chosen as the data model: firstly, it is a web standard for
how data should be published online, secondly RDF as a model (subject – predicate –
object) is easy to understand for most users, which is important when creating a gazet-
teer.
Historical place as a semantic web concept needs to address at least the following
four basic classes: place name, spatial location, source and time. The principles behind
the last three are discussed in the sections above. Regarding time, we have found that
the concept Historical settlement units are considered to be rather stable over time. For
this reason time is not an attribute of the data model, rather the units are considered to
have a potential validity extending from c. 1000–1800. Information about when a spe-
cific settlement appear in various sources could be linked to a settlement unit, and does
not need to be attributed in the core class definition. An attempt to define validity in
time for individual settlements would much likely fail. Information on starting and end-
ing rarely occur in the source material, and has to be based on subjective assumptions,
which is not suitable in basic register. Temporal attributes are likely better suited for
more general geographical data (as e.g. parishes).13
The usage of source in the data model specifically refers to the historical maps that
points out the coordinates. A separate class is used to store these data
(TORA:CoordinateSource), which holds references to the core class
(TORA:HistroicalSettlementUnit). In other words, the map(s) that have been used to
identify the location constitutes the coordinate source. The oldest map is primarily re-
ferred, but in many cases more than one map are relevant as a source.14
Place names that are associated with a particular settlement could also be managed
with the help of related classes. Term search related to place names is of course essential
for a gazetteer. But how to attribute a place name to a historical settlement unit is not
always obvious. A historical name of a settlement, known from sources, does not nec-
essarily correspond with the contemporary name for the same place. Different sources
can also provide different names. And it is often not reasonable to manage all occur-
rences of names when defining a settlement unit in a basic register. A concept like
Historical settlement unit need a preferred name, to be shown along with the coordi-
nates. The name to be used is not always given. It should be a name valid and relevant
13
A historical gazetteer with temporal attribute is proposed by i.e. http://whgazetteer.org.
14
The coordinates that are extracted from the historical maps have been stored using the Swe-
dish official coordinate system or projection, SWEREF 99 TM (EPSG:3006). Today's GIS
tool allows easy conversion between different coordinate projections. In order to simplify re-
use, both the national and international standard WGS 84 (EPSG:4326) are exposed as linked
data. However, when the coordinates are rendered on a Web page it is often preferable to use
the Mercator projection (EPSG:3857) which translates WGS 84 into a sphere.
265
for the period (in this case c. 1000–1800) that is documented in the sources. To exem-
plify, if a rural village at some time has developed into a town with a different name,
or if the name changed for other reasons, the historical name should be given primarily,
while the later represents alternative names.
The chosen way to manage different place names for a settlement is to choose a
preferred name (skos:preLabel), while all other names are handled in a separate related
class with alternative names (TORA:AlternativeName), that could be derived from dif-
ferent sources.
Another aspect of place name concerns spelling. It is accepted that names should be
spelled according to the usual spelling on present official maps. However, if names only
appear in older sources, it may be difficult to normalize the spelling. Then it may be
better to quote the name according to a source instead. The chosen way to handle this
have been to classify each name as either modern, normalized or cited in a related class
(TORA:NamingPrinciple), where “modern” relates to names appearing on official
maps, “normalized” refer to lost place names with a standalone spelling, and “cited”
refer to a quote in written source. Many historical names have a known name forms
(prefix/suffixes) and could therefore be normalized, which facilitates reading and use.
Another reason to normalize names is that older spellings vary – the same name can be
spelled in several ways in the same document.15
15
The naming principle used in the TORA project has been developed and discussion in coop-
eration with researchers at the Institute for Language and Folklore in Uppsala, and may be
further refined.
266
dcterms:references tora:CoordinateSource
dcterms:references tora:AlternativeName
https://
data.riksarkivet.se/ skos:prefLabel Gamla Uppsala @sv
tora/1335
dcterms:identifier 1335
wgs84_pos:lat 59,8945
wgs84_pos:long 17,6402
dcterms:coverage tora: CoordinateAccuracy
tora:ModernParish/
dcterms:isPartOf
03800003
tora:EarlyModernParish/
dcterms:isPartOf
038003000
rdfs:type tora:HistoricalSettlementUnit
Fig. 1. The Historical settlement units are exposed as linked data in the RDF format. Each unit
has a stable URI which constitute the subject or the resource. The relations are express as predi-
cates using standard vocabulary or TORA terms. The objects are either nodes or attributes with
a value string. The complete TORA model include other referenced data, but this figure focuses
on the core concept of Historical settlement unit.
267
Fig. 2. An overview of the current (February 2019) Historical settlement units registered in
TORA, c. 27.000 units covering large parts of Sweden, derived from large scale maps and his-
torical source before c. 1800.
6 Conclusion
A Historical settlements unit, as defined in this article, refers to villages, hamlets and
single farms which formed the building blocks in the rural society from the Middle
Ages until the early modern period. It intends to define historical places, which lack
clear definition on the Semantic Web today. It tries to meet the researchers need for
better historical context and data definitions, and uses linked data technologies for mod-
eling and publication.
268
In brief the following criteria constitute a Historical settlement unit: 1/An accurate
coordinate pair (latitude and longitude), that are 2/derived from historical map(s), pref-
erable large-scale survey maps; 3/a chosen place name that is vailed for the settlement
in its historical context, along with 4/a specified spelling principle and 5/alternative
names when so occurs.
This paper concludes that the simple RDF data format is sufficient when creating a
gazetteer, although historical settlement is a complex spatial object, that could be of
different kind (village, town, manor etc.) and hold temporal aspects and several place
names. The simplicity of the RDF is seen as a strength, which clarifies what constitutes
the core classes and its relations and metadata. In the TORA project, only spatial coor-
dinates as point objects (latitude and longitude) have been handled. If additional geo-
data is to be included, such as polygons with time intervals for administrative units such
as parishes and larger areas, the model likely needs to be expanded.
Different fields of research, such as history, archeology and onomastics, would ben-
efit from online historical geographical registers, for instance when analyzing relations
between people, events and places. It could also be used as a tool in community plan-
ning and be a resource in genealogy and local historical research.
7 Acknowledgement
This work presented here is a part of the TORA project, that has been initiated by a
group of researchers and co-workers at several institutions in Sweden, with funding
(2015–2019) from Kungl. Vitterhetsakademien, Riksbankens Jubileumsfond and the
Swedish National Archives. A first public launch will take place 2019
(https://riksarkivet.se/tora-english).
References
1. Aucott, P., Fleet C. and Southall, H. 'The GB1900 project – from the horse’s mouth', Sheet-
lines 111 (April 2018), https://s3.eu-west-2.amazonaws.com/sheetlines-articles/Is-
sue111page46.pdf
2. Braudel, F. Histoire et sciences sociales: La longue durée. Annales: Économies, Sociétés,
Civilisations, vol. 13, n. 4, pp 725–753 (1958).
3. Kain, Roger J. P. & Baigent, E. The Cadastral Map in the Service of the State: A History of
Property Mapping. Chicago: University of Chicago Press (1992)
4. Karsvall, O. Uninhabited cadastral units on large-scale maps 1630–1655: indicators of late
medieval deserted farms? In: Landscape History Vol. 29, 2007:1, pp 75—87 (2007),
https://www.tandfonline.com/doi/abs/10.1080/01433768.2007.10594590
5. Karsvall, O. Utjordar and the question of deserted farms − a case study of the parish of
Svanshals. In: Bebyggelsehistorisk tidskrift, vol. 61, 2011, pp 22–38 (2011), http://me-
dia.bebyggelsehistoria.org/pdf/BHT61_2011_22-38.pdf
6. Karsvall, O. Retrogressiv metod: En översikt med exempel från historisk geografi och agrar-
historia. In: Historisk tidskrift (Sweden) 133:3, pp 411−435 (2013), http://www.historisk-
tidskrift.se/fulltext/2013-3/HT_2013_3_411-435.htm
269
7. Karsvall, O. Utjordar och ödegårdar: En studie i retrogressiv metod. Diss. Uppsala (2016),
https://pub.epsilon.slu.se/13711/
8. Karsvall, O. & Borin, L. SDHK meets NER: Linking place names with medieval charters
and historical maps. In: DHN 2018, Digital Humanities in the Nordic Countries 3rd Confer-
ence. Mäkelä & Tuominen (eds), pp 38−50 (2018), http://ceur-ws.org/Vol-2084/paper3.pdf
9. Panecki, T., Związek, T. & Myrda, G. The Spatial Database Development for Research on
Settlements and Afforestation in Nowy Tomyśl Plain in Early Modern Times. In: Studia
Geohistorica, Nr 06. 2018, pp 186−205 (2018), http://atlasfontium.pl/index.php?arti-
cle=webhauland
10. Southall, H., Aucott, P., Fleet, C., Pert, T. and Stoner, M. ‘GB1900: Engaging the Public in
Very Large Scale Gazetteer Construction from the Ordnance Survey ‘County Series’
1:10,560 Mapping of Great Britain.’, Journal of Map & Geography Libraries, 13 (1), (2017),
pp 7–28
11. Wickham, C. Framing the Early Middle Ages: Europe and the Mediterranean 400−800. Ox-
ford: Oxford University Press (2005)