Professional Documents
Culture Documents
METHODOLOGICAL PERSPECTIVE
Du, Q.
School of Resource and Environmental Science, Wuhan University,
129 Luoyu Rd., 430072 Wuhan, P. R. China.
E-mail: qydu@telecarto.com
ABSTRACT
Geographic information is the most important component of the human cognition of reality, which is traditionally
represented by map. Besides verbal language and music, map is regarded as “the third language” for human beings
communication. In this paper, the author argues that the theory of map language can further provide Geographic
Information System with discipline paradigm and corresponding conceptual model while traditional map theories of
cognition, communication and language have been providing a sound theoretical foundation for Cartography, and a
completed descriptive structure of geographic information based linguistic paradigm is proposed in this paper.
1. INTRODUCTION
Linguistic paradigm is not new in geographic information related disciplines. The most important one is map language
theory that stemmed from Bertin’s retinal variables (Bertin, 1967) and one of the three most popular cartographic
theories. Map language paradigm regards map either as analog of natural language or semiotic system, by which the
association of content and express or referent, communication and interpreter is investigated (Schlichtmann, 1979-1999,
Head, 1999). The main concern of this approach is for the conceptual model of map and its digital form (Schlichtmann,
1999).
Another approach of linguistic paradigm is from the viewpoint of cognition which stemmed back from Russell’s idea;
the structure of language corresponds somehow to the structure of objective world. The typical example of this
approach is what was carried forward by the initiative 2 of NCGIA, linguistic aspect of spatial relations (Mark, 1988-
1992; Frank & Mark, 1991; Egenhofer, 1998).
There is also the third approach, which is still popular in the disciplines such as pattern recognition and digital image
processing and its descendent in cartography and GIS (Youngmann, 1978; Taketa, 1979; Nyerges, 1991; Du, 1997-
1998), which is more from the methodological perspective.
Although linguistic paradigms are widely accepted in the geographic information related circle, a genuine linguistics
oriented total investigation is absent, especially from the viewpoint of microcosmic level. The research of linguistic
paradigm in this paper is from the concern with methodology and in microcosmic level, which regards the
representation of geographic information as the analogy of natural language for the construct of the phonetic, semantic
and syntactic theory of geographic information as a two-dimensional graphic language.
These linguistic principles tell us how to construct a linguistic system, which embedded with widely utilized linguistic
terminology such as first of all Phonetics, Semantics, Lexics and Syntax, and is composed of a more complicated
internal structure in each element and component.
In Linguistics, Phonetics deals with the physical characteristics of language, which is the sound in verbal language and
“stroke” of character in writing language. In graphic representation of geographic information, then physical
characteristics are carried by graphic ink stroke, rather than sound, thus research of the physical characteristic of
graphics becomes the phonetic analysis of spatial information language.
Proceedings of the 21st International Cartographic Conference (ICC) Durban, South Africa, 10 – 16 August 2003
‘Cartographic Renaissance’ Hosted by The International Cartographic Association (ICA)
ISBN: 0-958-46093-0 Produced by: Document Transformation Technologies
1520
Dimensionality is another physical characteristic of spatial information. Point, line, area and volume are four
geometrical components with different dimensions. Among them point is indivisible and the others are extended
components which can be divided further.
As the representation of geographic reality, the appearance of geographic object is the whole presentation of physical
characteristics. In fact the appearance is dominated by three factors which include geographic reality, cognitive
regulations and geometric rules.
The phonetic syncopation based on retinal variables as Distinctive Feature (DF) is the most important procedure in
phonetic analysis. There are different syncopation methods stemmed from different disciplines and for different
purposes, among which are geometric, psychological syncopation in pattern recognition, mathematic morphology,
cartography.
Phoneme is the minimal phonetic meaningless element that functions as signaling the distinction of meaning in higher
language unit. In geographic information, retinal variables and dimensionality have the potential to act as Distinctive
Features to make the phoneme based syncopation. After analysis of eight retinal variables we can find that for phonetic
analysis they are still not atomic enough, among which only color, brightness, size and orientation are atomic features.
As to dimensionality, point and line segment are in two different dimensions, which can not be further divided (as
division of line segment results in line segment again). So point or line segment with variable color, brightness, size and
orientation form the phoneme of graphic language system. All other phonetic aspect of linguistic component is the
combination of these phonemes.
The power of graphic symbol expression comes from the combination of the phonemes, besides allophones of a certain
phoneme. The combination of phoneme can result in minimal meaningful unit (morpheme), or still meaningless unit
(syllable). In the background of geographic information, phoneme combination is traditionally the domain of map
symbolization.
3. SEMANTICS
Semantics deals with the meaning of language, which in geographic information the association of geographic
information and geographic reality. There are two main semantic theories, one of them is concerned with semantics
inside of language system, i. e. how words and sentences are mutually connected, another of them is concerned with
semantics outside of language system, i. e. how word and sentence are connected with their referent object and process.
So far in Linguistics efficient theory for semantic feature is lacking, empiricism and introspection are two main means.
As to geographic information, the corresponding geographic ontology is much confined comparing with reality
corresponding natural language. By the means of ontological investigation is an efficient way.
Borgo et al.(1996) and Guarino(1997) proposed the concept of “Ontological Strata” (Fig. 1) for the construction of large
scale ontology.
Static (a situation)
Mereological (an amount of matter)
Physical
Topological (a piece of matter)
Morphological (a cubic block)
Functional (an artifact)
Biological (a human body)
Intentional (a person or a robot)
Social (a company)
1521
“Objects have an intensional criterion of identity, in the sense that they are more that mere sums of parts. Within
objects, further distinctions can be made according to the identity criteria ascribed to them.” As for geographic information,
static, biological strata have less influence, mereological and physical strata not only have effect on internal semantic
features, and also function over whole structure of semantics.
As an example, we can make following decomposition into semantic features of geographic information.
! River——[water] + [flowing] + [naturally curved] + [transportation] + [linear];
! Lake——[water] + [standing] + [tourism] + [breeding] + [area];
! Highway——[artificial material] + [transportation] + [constraint curved] + [linear] + [economic];
! Fence——[artificial material] + [obstruction] + [regular] + [linear];
! Building——[artificial material] + [inhabitancy] + [regular appearance] + [area] + [obstruction] + [political and
economic meaning];
Of course, using only identifying criteria of everyday world object for geographic entity and phenomenon has its
shortage. For geographic concepts such as bay and cape or political boundaries (Smith & Mark, 1998), we still find its
semantic features from syntactic framework.
According the argument, semantic relations fall into four categories, i. E., synonymy, hyponymy, meronymy and
antonymy (Buitelaar, 2001).
3.2.1 Synonymy
Synonymy refers to the same or similar meaning of the words. In the representation of geographic information we use
strictly artificial language system. Represented object are classified and upon them the lexical system is prescribed so
that strict synonymy does not exist in it. However, if we look closer to the potential lexical system of geographic
information (both analog and digital), similar semantics with multiple representation is very common.
It includes:
! Multiple expression of the same entity: the variation of shape, color, weight result the multiple expression of the
same entity.
! Multiple expression of the same spatial information: for example, the different expression of the cross of highway.
3.2.2 Hyponymy
Hyponymy refers to also relation of similarity, but on the similarity of class. It is the inclusion of classes. For example,
vehicle includes motor vehicle and non-motor vehicle.
It can be said that the deep foundation for hyponymy of word is the hierarchical structure of concept. For geographic
ontology itself is a hierarchical structure(Smith et al., 1998; Mark et al., 1999), hyponymy has reason to be an important
relation in the lexical relation of geographic information. In terms of expression of geographic concept, phonetics,
lexics and syntax have the effect of constraint. The actual category of language unit is much less than that of geographic
concept so that they are not the relation of one-to-one.
The difference is the essence of representation by which infinite concept can be expressed by finite symbols.
1522
3.2.3 Meronymy
Meronymy refers to the part–whole relation of object or object class represented by words. In linguistic structure of
geographic information, meronymy is especially important for any geographic feature is essentially compound one
consisting of many parts. Smith & Mark(1998, 1999) regard Mereology as one of three basic tools for ontological
investigation of geographic kind, and Mereology is actually the main concern of formal ontology in Knowledge
Engineering (Simons 1987;Guarino 1995).
While Hyponymy has epistemic attribute, Meronymy has strong ontological attribute. Meronymy is more dominated by
the internal physical rules of reality.
3.2.4 Antonymy
Antonymy refers to the contrary relation of objects. There are different category of antonymy such as gradable
antonym, binary antonym and relational antonym. Antonymy is also reflected in geographic information as follows:
! Morphology-driven Antonymy – curve/straight, regular/irregular
! Class-driven Antonymy – Block/street, sea/land, mountain/plain, valley/ridge, urban/countryside
! Attribute driven Antonymy – Highway/auto way, Long-standing river/Seasonal river
! Relation driven Antonymy – above/below, overlap/overlapped, contain/contained
4. SYNTAX
Syntax is a very important concept, either in the view of linguistic system such as “phonetics-lexics-syntax”,
“phonetics-syntax-semantics” or “syntax-semantics- pragmatics”, where syntax is always at the heart of the system. The
function of syntax is to integrate language units with different physical features and conceptual meaning into higher
language unit which conforms to regulations and conveys a certain meaning.
Just as phonetic and semantic structure, syntax has its internal structure. In syntax (grammar), lexis and syntax are two
different units, which function in building of word and sentence respectively.
What kind of spatial relations exist between spatial entities is always a hot topic in Geographic Information Science
(Egenhofer, 1989). Following are some binary relations which reveal some combination relation from respective angles.
Topological relations are the most stable and yet most important ones in geographic features so that they are very easy
to cognize and store in computation system. However, based on the following reasons, more combination relation of
different types must be investigated.
! Firstly, topological relations are loose relations whose requirement is very easy to meet. By the further refinement
of spatial relation we must seek other type of relation. Some word can be defined only in the whole context.
! Secondly, disjoint relation take the most part of spatial relation in the real world. For those disjoint entities we need
also other type of relation.
! Thirdly, as a highly formalized description, topological relations need semantic and ontological constraint in
pragmatic context.
1523
4.1.2 Metrical Relation
If topological relation can be said having been thoroughly investigated, we have no much knowledge about what be
roughly called metrical relation.
Metric and topological are not two contrary concepts, but two aspect of the same spatially distributed phenomenon.
Different forms of topological and metric relation hold between any two spatial entities. Both of them can be computed
from their positions. Topological relation gives its qualitative aspect and metric relation gives other aspect of quantity.
The only way to describe metric relation is to approximate graphics as accurate as possible in analog representation.
Direction relation and distance relation can be defined between any two points, lines and areas. Because metric relation
is difficult to process in a qualitative way like topological one, metric relation is neglected in most Geographic
Information research. Some more complicated definition comes from qualitativation of direction and distance relation
for the sake of simulating human’s qualitative reasoning (Peuquet and. Zhan, 1987; Papadias and Sellis, 1994).
Here we find topological play a less important role than pattern of buildings. Metric relation is more obvious here. For
further definition of spatial combination of geographic entities, we need further refinement of our spatial information
theories, first of all making metric relation a qualitative one.
Some scholar has pointed out topological relation in geographic domain is not genuine topological in the sense of
topology (Smith and Varzi, 1997). In fact the most common examples of them are shifting, rotating and scaling, and
map projections are also included, but most of the projections are below second-order transformation.
If we hold suppose:
! Fist-order transformation is limited in shifting, rotating and scaling, and,
! map projection is blow second order, and spatial scale is relatively small;
where d12, d34 are distance between point 1 to point 2 and between point 3 to point 4 respectively
Definition 2: Da = |a12-a34|
where a12, a34 are azimuth between point 1 to point 2 and between point 3 to point 4 respectively will remain unchanged
basically. We can define an intermediate geometry, called Combinational Qualitative Geometry (CQG), based on the
two invariants, ratio of length and difference of azimuth. From the two atomic properties, more qualitative aspects of
metric relation can be deduced.
1524
Equal-length relation is a very usual combinational relation in geographic space. Its ontological basis is most spatial
features esp. artificial features have their statistically invariant magnitude range, such as for breadth of street, length of
river branch. In spatial association the occurrence of equal-length relation will increase dramatically.
Definition 7: For set of line segments S={o1,o2,…, on | n≥3}, if at least one oj∈S makes oi (Pa ◦ Ml)oj for each oi
∈S, S is called a Line Co-Linearity Set. Any number of elements above two have Line Co-Linearity
Relation, noted as Cl.
Definition 8: For set of points P={p1, p2, …, pn | n≥3}, if the set S={li,j | 1≤i<n, 1≤j<n } composed of the connection
of the any two elements of set P is a Line Co-Linearity Set, then P is called a Point Co-Linearity Set.
Any number of elements above two have Point Co-Linearity Relation, noted as Cp.
Definition 9: For any sequence so={ oj | oj∈S } by the subset of a Co-Linearity Set S, if oj(Pa ◦ Ml)oj+1 and Xoj ◊ X
oj+1 (where ◊ means “>” or “<”) hold for each element oj and its successive element oj+1 , so is called a
Sequence Relation based on set S, noted as Sp. Meanwhile, Nd = {(oj , oj+1)} is Direct Neighborhood
Relation.
CQRs based on metric relation could be regarded as the atomic spatial relations on extended geographic feature, which
are immediately applicable to some artificial and natural features such as block and pipe system, but may vary with the
most natural features with real geographic configuration. CQRs are to loose their constraint further.
1525
So far most GISs depend heavily on Euclidean Geometry. However, much research reveal that the structure of
geographic information is not merely a geometric matter, instead, more an ontological, epistemic and (natural) linguistic
matter. Spatial relation provides us with syntactic rules, which can establish linguistic structure meaningful to convey
real and meaningful information only integrated with phonetics and semantics.
The representation of spatial relation begins at the stage of phonetics. At the stage of Superasegment, based on curve
and minimal simple polygon, we can define more spatial relation which are the phonetic anamorphosis of both
topographic relations and CQRs such as the sequence of same level of curves, containing of different level of curves
and contacting of multiple level of minimal simple polygon as in double-line river system. Especially CQRs will loose
their phonetic constraints to accommodate the situation of natural distribution, where curve based parallel,
perpendicularity relation and equal-length relation can be defined.
In sentence of type 1), word is connected with reference system. It reflects a spatial situation. In sentence of type 2),
spatial relation of two objects with different semantics is given; here relative position is more important than absolute
position. In sentence of type 3), spatial relation of two objects with the same semantics is given. As a different point,
objects in this sentence are liable to merged into a new language unit when some condition is satisfied as neighborhood.
In this case spatial relation tends to be constraint between language units instead of acts as verb.
On the other hand, semantics of spatial information has its ontological basis. By linguistic approaches we can
investigate ontological semantics based on the participants of relations as following:
! Dimensionality property: A certain spatial relation has requirement of dimensionality for participants. As an
example, containing relation requires a area object as container;
! Active/passive: Some spatial relation implies the relation of active/passive. For example, line may be active than
area in line/area relation as road gets through park.
! Vertical relation: although two-dimensional spatial relations are defined on plain, participants are not always on a
plain. Following situations are possible:
! On one plain: as road and farm field;
! Above: as bridge and river
! Below: as tunnel and mountain
! Uncertain: as highways at cross.
! Compatibility: some spatial relations require the compatibility of the participants, otherwise is impossible in reality.
For example, conceptual object such as political boundaries can share location with river while highway can’t.
! Spatial constraint property: some spatial relations imply the spatial constraint and correlation such as parallel
relation.
! Cause property: Such spatial relation as co-linearity, equal-length implies artificial construction. Some certain
pattern implies certain natural cause etc.
1526
Table 1. Semantic anamorphosis of the same spatial relation in different contex
Spatial Relation Participant Semantic anamorphosis
Road, park Go through
Bridge, river Cross over
Line/area
Dam, river Block the water of
Overlapping
Ferry course, river Get across the surface of
Waterfall, river Fall as the part of
Reef, sea Submerging in
Area/point Elevation point, road On the surface of
containing Buoy, river Floating on
Cave, mountain In
The semantic feature of geographic information is the prerequisite for our better understanding of geographic space.
They are linguistic knowledge standing apart from concrete geographic configuration. Table 1 gives an illustration of
how certain spatial relation obtain their semantic anamorphosis in ontological context.
5. CONCLUSION
As in this paper, even mere from the microcosmical and technological approach, we can see that linguistic paradigm of
geographic information has at least four aspects of significance for the development of GIScience, i.e., its paradigm
potential, its ontological concern, its methodological guidance and its qualitative approach.
In the framework of linguistic paradigm, we have carried out some technical practice on map symbol recognition (Du,
1998) and the linguistic analysis of multimedia electronic atlas (Du, 2001). The further research will include the
investigation of how the physical features and its ontological knowledge could be integrated into the spatial relations to
enhance the automatic understanding ability of geographic information system and how it can benefit spatial data
mining and knowledge discovering. On the other hand, the possibility of an investigation of map semiotic system across
spatial, temporal and cultural coverage with the aim to reveal the evolution of human spatial cognition is also within our
sight.
6. ACKNOWLEDGEMENT
The author would like to acknowledge the support of Natural Scientific Foundation of China under project No.
40071071, No. 49601015 and Hok Ying Tung Education Foundation for the research.
7. REFERENCES
[1] Bertin, J. 1979. Visual Perception and Cartographic Transcription, World Cartography 15: 17-27
[2] Borgo, S., Guarino, N., and Masolo, C. 1996. Stratified Ontologies: the Case of Physical Objects. In:
Proceedings of ECAI-96 Workshop on Ontological Engineering. Budapest.
[3] Buitelaar, P. 2001. Lexical Semantics in Information Systems, http://dfki.de/~paulb/ lexsem/2001/lexsem.html
[4] Du, Q. Y. 1997. Linguistic characteristics and automatic understanding of cartographic information. In:
Proceeding of the 18th International Cartographic Conference, Stockholm, 357-364
[5] Du, Q. Y. 1998. Map language based cartographic understanding, in: Proceeding of ’98 ACSM Annual
Conference and Exhibition, Baltimore, 253-257
[6] Du, Q. Y. 2001. From Micro to Macro: An Exploratory into the Structure of Multimedia Electronic Atlas. In:
Proceeding of the 20 International Cartographic Conference, Beijing
[7] Egenhofer, M. 1989. A Formal Definition of Binary Topological Relationships, in: W. Litwin and H. Schek
(eds.), Third International Conference on Foundations of Data Organization and Algorithms (FODO), Lecture
Notes in Computer Science, Paris, France, Vol. 367, Springer-Verlag, 457-472
[8] Egenhofer, M. and Franzosa, R. 1991. Point-set topological spatial relations. International Journal of
Geographic Information Systems, 5:161-174
[9] Egenhofer, M. J., and Mark, D. M., 1995. “Naive Geography,” in Frank and Kuhn (eds.), European Conference
on Spatial Information Theory Lecture Notes in Computer Science, Semmering, Austria, Vol. 988, Springer-
Verlag, 1-15
[10] Egenhofer, M. and Shariff, R. 1998. Metric Details for Natural-Language Spatial Relations. ACM Transactions
on Information Systems, 16
[11] Guarino, N. 1995. Formal Ontology, Conceptual Analysis and Knowledge Representation. International Journal
of Human and Computer Studies, 43(5/6): 625-640.
[12] Guarino, N. 1997. Some Organization Principle For A Unified Top-Level Ontology, in: Working Notes of AAAI
Spring Symposium on Ontological Engineering, Stanford
1527
[13] Head, C. G. 1999. Warp and Woof: Carto-semiotics and Carto-linguistics in the English-language Literature, in:
Map Semiotics around the World, ICA Publication
[14] Mark, D. M. 1988. Cognitive and Linguistic Aspects of Geographic Space: Report on a Workshop. NCGIA
Technical Paper 88-3
[15] Mark, D. M. et al. 1989. Languages of Spatial Relations: Initiative Two Specialist Meeting Report, Santa
Barbara, CA: National Center for Geographic Information and Analysis, Report 89-2.
[16] Mark, D. M. 1991.Spatial Representation: A Cognitive View, In: Geographic Information System: Principles
and Applications, edited by David Maguire et al., New York, NY: Longman Scientific & Technical
[17] Mark, D. M. and Frank, A.U.1992. NCGIA Initiative 2 “Languages of Spatial Relations” Closing Report.
[18] Mark, D. M., Smith, B., and Tversky, B. 1999. Ontology and Geographic Objects: An Empirical Study of
Cognitive Categorization. In : Freksa, C., and Mark, D. M., editors, Spatial Information Theory: A Theoretical
Basis for GIS, Berlin: Springer-Verlag, Lecture Notes in Computer Sciences, 283-298
[19] Nyerges, T. L. 1991. Geographic Information Abstractions: Conceptual Clarity for Geographic Modeling.
Environment and Planning A 23, 1483-1499
[20] Papadias, D. and Sellis, T. 1994. On the Qualitative Representation of Spatial Knowledge in Two-Dimensional
Space. Very Large Data Bases Journal., VLDB Special Issue on Spatial Databases
[21] Peuquet, D. and. Zhan C.-X. 1987. An Algorithm to Determine the Directional Relationship between Arbitrarily-
shaped Polygons in a Plane. Pattern Recognition 20 : 65-74.
[22] Schlichtmann, H. 1979. Codes in Map Communication. The Canadian Cartographer 16/1: 81-97
[23] Schlichtmann, H. 1985. Characteristic Traits of the Semiotic System ‘Map Symbolism’, Cartographic Journal
22/1: 29-30
[24] Schlichtmann, H. 1994. Map Symbolism Revisited – Units, Order, and Contents, in: Pravda J. et al. 1994.
Cartographic Thinking and Map Semiotics, Geographia Slovaca 5, 47-62
[25] Schlichtmann, H. 1999. An Inventory of Research in Map Semiotics. In: Map Semiotics around the World, ICA
Publication
[26] Shariff, A. R. and Egenhofer, M. and Mark D. 1998. Natural-Language Spatial Relations Between Linear and
Areal Objects: The Topology and Metric of English-Language Terms, Internationa Journal of Geographical
Information Science, 12/3:215-246
[27] Simons, P. M., 1987. Part: An Essay in Ontology. Oxford: Clarendon Press.
[28] Smith, B., and Varzi, A., 1997. Fiat and Bona Fide Boundaries: An Essay on the Foundations of Geography. In:
S. C. Hirtle and A. U. Frank (eds.), Spatial Information Theory. International Conference COSIT '97. Laurel
Highlands, Pennsylvania, (Lecture Notes in Computer Science 1329), Berlin/New York: Springer Verlag,
103-119.
[29] Smith, B. 1998. Basic Tools of Formal Ontology. In N. Guarino (ed.) Formal Ontology in Information Systems.
IOS Press
[30] Smith, B. & Mark, D. M. 1998 Ontology and Geographic Kinds. Proceedings of SDH ’98, Vancouver, Canada
[31] Taketa, R. A. 1979. Structure and Meaning in Map Generalization. Unpublished PhD dissertation, University of
Washington, Seattle, WA, U. S. A.
[32] Youngmann, C. E., 1978. A Linguistic Approach to Map Description. First International Advanced Study
Symposium on Topological Data Structure for GIS. Vol. 7. Harvard University
1528
A CARTO-LINGUISTIC PARADIGM TAKING A
METHODOLOGICAL PERSPECTIVE
Du, Q.
School of Resource and Environmental Science, Wuhan University,
129 Luoyu Rd., 430072 Wuhan, P. R. China.
E-mail: qydu@telecarto.com
Biography
Dr Qingyun Du is Professor of Cartography and GIS, and deputy director of School of Resource and Environmental
Science at Wuhan University. His research interests include linguistic conceptual modal of geographical information,
automatic cartographic generalization, electronic mapping and spatial visualization. He serves on 2 editorial boards for
journals. He is corresponding member of ICA Commission on Theoretical Cartography and co-chairs for two Chinese
association commissions on GIS and Cartography. He has lead more than 30 projects and received 8 national and
ministerial scientific awards.
1529