Place Name Ambiguities in Urban Planning Domain Ontology

Yi-Pei Liao

and Feng-Tyan Lin

1,2

Department of Urban Planning, National Cheng Kung University, Tainan, Taiwan

College of International and Cross-Strait Education, Asia University, Taichung, Taiwan

Keywords: Urban Planning Domain Ontology, Spatial Ontology, Place Name Ambiguity.

Abstract: Integrating spatial information is a crucial step in construction of Urban Planning Domain Ontology

(UPDO), and taking spatial information from the web as the input of self-learning method are commonly

used in constructing ontology. In this case, the use of place names could be important indicators of

understanding the spatial information on the web. However, place names expressed in natural language

bring diverse ambiguities, which would bring great challenges to several research fields such as Geographic

Information System (GIS) and Geographic Information Retrieval (GIR). GIR has more contribution on

place name ambiguities than GIS. Nevertheless, from the perspective of the urban planning domain, it is still

lacking in application. This paper is a position paper that aims to bring out an argument of place name

ambiguities in UPDO, and introduce two kinds of ambiguity frequently appearing in the urban planning

domain. The paper also proposes a hierarchical structure of spatial ontology that allows constructors to deal

with ambiguities. We believe in that the ambiguity issue is critical for urban planning, and the argument is

worth discussing to all relevant domains.

1 INTRODUCTION

In construction of Urban Planning Domain Ontology

(UPDO), the integration of geographic information

undoubtedly is an important task. In practice, UPDO

needs spatial information rather than geographic

information as spatial information involves

connections of locations, people, and activities (Lin

et al., 2013). In view of that, place names could play

an important role in integrating spatial information

especially when ontology constructors need to

retrieve spatial information from the web. However,

more and more place names have been created

rapidly and informally through various internet

activities, for instance, tagging function on

community websites. Matching a place name to a

real space therefore turns more ambiguous than

before as a place name is more likely to become a

vague concept instead of a geographic information.

Moreover, some of conceptual place names are

widely used in formal documents, such as press

releases and formal governmental reports. This

brings out a new challenge for UPDO to deal with

the ambiguity.

In fact, place name ambiguities are not a new

issue. The research field of Geographic Information

System (GIS) has studied on this question since the

1990s, and so did the field of Geographic

Information Retrieval (GIR), a technique combining

the information retrieval and spatial ontology. GIR

has contributed to derive the geographic information

from web documents automatically, as a result,

recognizing the place names from natural language

is one of the great challenges in GIR. Nevertheless,

the existing disambiguating technique in urban

planning domain remains insufficient. More efforts

are needed to be put for solving place names

ambiguities while constructing UPDO which is

grounded on robust spatial ontology.

This paper is a position paper bringing out an

argument over place name ambiguities. It presents

two kinds of ambiguity that are commonly seen in

urban planning domain, yet have not been taken into

consideration in GIR. In addition, it proposes a new

hierarchical structure of spatial ontology which is

able to deal with the ambiguity. The paper is

organized as follows. Section 2 will review works

with regards to place ambiguities in GIS and GIR.

Section 3 introduces two kinds of place name

ambiguity that have not been dealt adequately in

terms of urban planning practices. Section 4 brings

out a new structure of spatial ontology which

considers the place name ambiguity discussed in

Liao, Y. and Lin, F..

Place Name Ambiguities in Urban Planning Domain Ontology.

In Proceedings of the 7th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2015) - Volume 2: KEOD, pages 429-434

ISBN: 978-989-758-158-8

429

Section 3, and proposes a brief outline for

proceeding to construct the spatial ontology. Finally,

conclusion is given in Section 5.

2 BACKGROUND

In the following are reviews of works regarding

place name ambiguities. The first part will firstly

focus on the field of GIS, and then shift the focus to

the spatial ontology and GIR.

2.1 Place Ambiguities in GIS

Name of place is a basic attribute of location

information in common spatial database,

accompanied with other basic attributes like latitude,

longitude, altitude, coordinates, area, etc. Most of

spatial databases advocate an entity-oriented view of

space, which means that the space data which could

be in the form of a point, a line, or an region is an

exact object. Some difficulties are discovered when

it comes to defining an area with indeterminate

boundaries (Burrough and Frank, 1996; Wang and

Hall, 1996). One of difficulties is a vague region or

fuzzy spatial data types (Erwig and Schneider, 1997;

Schneider, 2008). Fortunately, many researches and

tools have advanced in dealing with the problem of

indeterminate boundaries and vague regions.

Fuzzy region problem has been handled in GIS,

however, the spatial data representing through the

open-data and internet platform is still based on the

traditional view of space, for example, Google Map

and Open Street Map (OSM). Furthermore, many

map tools commonly used by users and netizens

display the place data through a standard format,

such as KML or GML, yet the simplicity will bring

difficulties in recognizing the place from expression.

2.2 Place Ambiguity in Spatial

Ontology and GIR

Despite of the advancement in GIS, spatial ontology

development and knowledge management in dealing

the ambiguity problem remains in a primary state.

Numerous works have put efforts on constructing

ontology, but few of them have focused on the time

and space dimension of thematic ontology. Peuquet

(2001) worked on an ontology framework which

could derive effectively the what/when/where

information with robust space-time data structure.

Also, he developed the query language, the

operation, and the users interface. Perry et al. (2006)

drew an outline of basic classes and relationships for

a spatial upper-ontology, which brought spatial

dimension into other ontology and allowed spatial

query operation.

The technique of Geographic Information

Retrieval (GIR) has focused on the spatial relations

between any kinds of knowledge, where information

is described with geographic metadata. GIR is user-

oriented applications including spatial query, search

and display functions. For example, the Spatially-

Aware Information Retrieval on the Internet

(SPIRIT) (Jones et al., 2002; Jones et al., 2004) is a

search engine for geographic information. SPIRIT

advocates that most of the web resources refer to

geographic space, which means that the event was

recorded once when it reveals or happens in a certain

place. SPIRIT regards each entity as a geography

entity with geographic information. In the part of

query method, SPIRIT intelligently understands the

users’ searching language and tells any possible

event that has relation with geographic information.

Mata and Claramunt (2011) rested on the

contribution of SPIRIT and gave an approach for

retrieval of geographic entities according to its

spatial, temporal, and thematic information. The

approach extracted diverse dimension of information

from the gazetteer, which has its own XML format

of geographical entity (eg. Wikipedia).

Several challenges in GIR demand further

significant researches (Jones and Purves, 2008),

including (1) detecting geographic reference in the

form of place names and spatial natural language, (2)

disambiguating the place names, (3) indexing

documents respecting to their geographic context, (4)

ranking relevant documents with respect to

geography as well as theme, (5) developing effective

user interfaces, and (6) developing methods to

evaluate the success of GIR. The research

introducing in this paper focuses on the first two

challenges, the place names detecting and place

names disambiguating. This research also faces

another difficult but critical challenge of the

language characteristics in Chinese.

In summary, GIS and GIR are both very

fundamental tools in the domain of urban planning.

In GIS, there is advancement in representing places

even though data might be either specific or vague.

The problem of vague region is much like a

semantic problem of space, and ontology is

considered as a major method to deal with the

semantic problem. However, it is still in a primary

stage in terms of spatial ontology and GIR.

KEOD 2015 - 7th International Conference on Knowledge Engineering and Ontology Development

430

3 PLACE NAME AMBIGUITY

The nature language problem in different languages

is important to GIR, and the place name ambiguities

is one of top priorities needed to be dealt with, and a

lot of researches have put efforts on it. Nevertheless,

only a few of place names ambiguity researches

have dealt with the problem that would happen in

using Chinese language. Chinese characters,

different from English word, is defined separately by

intrinsic meaning rather than by visible space,

resulting in that a character cannot be recognized by

a normal pattern. Furthermore, in Chinese the

meaning of a character could become very different

when combining with other characters. Thus,

defining a character and a term is a critical task in

Chinese as its meanings are strongly dependant on

the grammar analysis system. The follows are going

to introduce two kinds of Chinese place name

ambiguity that often appear in urban planning

domain.

3.1 The Place Name as a Concept of

Spatial Distribution

A planner may need to query for location

information of specific spatial events, such as

“flooding area” and “potential flooding area”. In this

case, there are at least two ambiguity problems that

make “flooding area” unable to be identified: (1)

how can the character “flooding” be recognized as a

part of a place name? It is more likely to be

identified as an adjective. How to distinguish an

adjective for a place name from an adjective in

normal would then become a problem. (2) The term

“flooding area” refers to a concept of geographic

distribution instead of a certain location. However,

according to the assumptions of normal GIR, there

could be only one focusing place in a single place

name entity, hence “flooding area” is unrecognized

because it is a description of a spatial event in name

yet without a continuous distribution in space.

3.2 The Place Name as a Concept of

Social Phenomenon in Space

Another place name ambiguity is caused by

conceptualized social phenomena. This kind of place

names could be a proper noun individually that is

easy to be recognized by pattern. Nevertheless, it

also has a very fuzzy boundary or even has no

boundaries on a map. The reason is that these place

names are created for describing particular social

development phenomena such as poverty gap and

real estate. In other words, the phrases have been

named before being indicated in space. In addition,

these place names could be unprofessionally defined

as they may be created by the public, especially the

netizens.

One of typical examples for this type of

ambiguity is a compound term called “Tyan-Long

Nation”, which is firstly created by some netizens

and is now frequently used by news media in

Taiwan. “Tyan-Long Nation” has an ironic meaning

initially that indicates a place where its residents are

self-centered and ignorant about anything happened

in other places. Yet nowadays the meaning has

evolved into a phrase describing a place with high

commodity price and extreme high housing price.

Moreover, verbally “Tyan-Long Nation” refers to

the Taipei City, the capital of Taiwan, however, they

are not exactly matched geographically.

In planning domain ontology, it is necessary to

understand this type of place names. Remind that it

is probably hard to find “Tyan-Long Nation” on the

Google Map because it is a phrase for a specific

concept rather than a location. Furthermore, it is a

name created by the public rather than by the official.

Therefore, in order to reveal the fuzzy boundary of

“Tyan-Long Nation”, we collect all locations with a

name “Tyan-Long Nation” from the Facebook

places, in which locations are allowed to be created

by all users, and check how much Facebook users

have checked in at each location. The more the

location has been checked, the higher score as well

as the possibility is assigned to the location. Figure 1

shows the result of this survey. The boldest black

line is the boundary of Taipei City, and the grey

circles are all locations that are named with “Tyan-

Long Nation” each circle has the radius about 1.2

km, and with shade degree base on the number of

Figure 1: A map displays the “Tyan-Long Nation”

according to the mass Facebook users.

Place Name Ambiguities in Urban Planning Domain Ontology

431

checked users. Figure 1 clearly reveals that “Tyan-

Long Nation” is concerned with some places in

Taipei City but not all. In this survey, the place

name was created basing on the social phenomenon

with a vague region, and we display the region on a

map according to the diverse understandings of mass

Facebook users.

4 SPATIAL ONTOLOGY FOR

PLANNING DOMAIN

In order to develop a spatial ontology for urban

planning GIR and UPDO, we need a new structure

of spatial ontology that is able to handle the

ambiguity problems identified in Section 3. In the

following part will firstly introduce the source of

place names, then a new structure for spatial

ontology, and finally a framework of semi-automatic

construction of this spatial ontology.

4.1 Place Names from Facebook

Facebook is the major sources of place name

collection in this research. The contents are

generated from the users’ perspectives, thus how to

take these informal contents into scientific and

theoretical research becomes a critical point, which

could be a very fundamental question. In the case of

collecting place names from Facebook, it’s obvious

that not every check-in name could be regarded as a

place name, and in contrary, it’s also expected that

some informal check-in name should be identified as

a place name. The critical and fundamental question

is “what is a place name?”

Cresswell (1996) aimed that the concept of place

should be scrutinized to both geography and human

everyday life. A place name is created only when the

place have some relations with some human

activities. In the domain of urban planning, the

rethinking of the relation between space, human, and

symbols (names) is quite associated to the concept

of “City image,” which was brought out by Lynch

(1960). City image argues that a city’s space is not

defined by its structure design but by the feeling of

the people living in it. City image is a mental map of

a person, and it might be very different to each

person even though they are experiencing in the

same city. Based on the concepts above, it’s

interesting to look through the Facebook place

names and analysis the reason of why the name is

created, is it related to “what people feel about the

place?” or “what people do at the place?”

Figure 2 shows the procedure of Facebook place

names Extraction. There are four parts between the

extraction from web and the storing to database: (1)

API is a tool to collect data from Facebook by using

the Graph API. So far there are about 10,000 place

names in Taiwan have already collected from

Facebook. The metadata has 10 items: name,

category, street, city, state, country, zip, latitude,

longitude, and check-in. The check-in is the count

number of users who checked in with that place

name. (2) Potential place name extraction is working

on analysing the name and deciding whether it is

possible place name or not. There are several kinds

of situation that make the name none potential, for

example, there’s a name called “on a moving train”

or “car racing.” None potential place names are

regarded as unidentifiable names which will leave to

(3) place name re-identification process. In this

process, several algorithms are developed for

different unidentified situations. (4) Place name

formalization is the last step before storage.

Formalization will deal the structural ambiguity

problems, such as shortened names and alternated

names. It’s based on the previous work by Deng et al.

(2012), which has developed an algorithm to extract

Chinese place name by using natural language

processing (NLP) method.

Figure 2: The procedure of Facebook place name

extraction.

4.2 A Structure of Spatial Ontology

An ontology-based model is utilized in the new

defined spatial ontology. Figure 3 shows the UML

class diagram of the hierarchical spatial ontology,

which allows the situations mentioned in Section 3.

The place is divided into three levels. The place

class at the first level refers to a general concept.

Following the place class, normal place and

distribution area, stand at the second level. The

normal place bases on the prototype and algorithm

in previous research of GIR. The distribution area is

separated in two sub-categories, namely, spatial

event and social event, which are both learned from

KEOD 2015 - 7th International Conference on Knowledge Engineering and Ontology Development

432

UPDO. The distribution area could have

distribution rate relations with several normal

places, where each distribution rate describes the

frequency of the distribution area has distributed in

that normal place. The distribution rate relation has

a value Di, 0≦i≦n, where n is the number of total

normal place that have distribution rate relations

with that distribution area.

Figure 3: The class diagram of spatial ontology.

Figure 4: Instance diagrams of two kinds of distribution

area, where “flooding area” is a spatial event and “Tyan-

Long Nation” is a social event.

Figure 4 are instance diagrams of two examples

mentioned in Section 3.1 and 3.2. The “flooding area”

is a spatial event and “Tyan-Long Nation” is a social

event, yet both of them are belong to distribution

area. “Flooding area” distributed in “Yunlin”,

“Chiayi”, “Tainan”, “Kaohsiung”, and “Pingtung”,

while “Tyan-Long Nation” is dispersed to “Taipei

City”, “New Taipei City”, “Xinyi District”, “Da’an

District”, and “Shilin District”. The distribution

relations are recorded respectively as symbols D

. These distribution rate values can be calculated

by spatial analysis methods such as proportion of

area, or by textual analysis methods such as co-

occurrence rate, or by integrating the former two

methods.

4.3 Semi-automatic Spatial Ontology

Constructing

Based on the class hierarchy of place in Figure 3, we

construct a spatial ontology with three parts, the

normal place database, the distribution area database,

and the distribution relations database. Figure 5 is

an outline of semi-automatic construction procedure

of spatial ontology, in which the rectangle with bold

dotted line describes the spatial ontology.

Figure 5: Semi-automatic spatial ontology construction

framework in UPDO.

Figure 5 shows the procedure of semi-

automatically constructing spatial ontology, and

there are three points worth being mentioning: (1) A

Chinese grammar analysis tool called CKIP,

developed by Academia Sinica in Taiwan, would

fragment terms into sub-terms according to their part

of speech (Ma and Chen, 2003). CKIP involves in

the process of determining the category of place

from the first level to the second level. (2) The

existing UPDO is able to assist in identifying any

relevant spatial or social event that are embedded in

the place name, and thus would help to decide the

category of distribution area from the second level to

the third level of spatial ontology. (3) The

distribution rate is calculated according to both

spatial analysis and textual analysis methods.

5 CONCLUSION

This paper is a position paper that brings out an

Place Name Ambiguities in Urban Planning Domain Ontology

433

argument of place name ambiguities. Although some

ambiguities have been taken into account in GIS and

GIR researches, the domain of urban planning,

which especially needs to integrate all knowledge

existing in a particular space, still lacks for the

consideration. Two types of ambiguities in Section 3

are the evidences indicating a wide gap between the

place name and the physical space. We also propose

a new defined structure of spatial ontology that will

be utilized in UPDO in further researches. The

spatial ontology presented in Section 4 is a

fundamental framework for urban planning GIR and

UPDO. We believe that the contribution of this

research in further can serve in several tasks such as

Decision Support System (DSS), knowledge

understanding, and the automatic learning of

relevant domain ontology.

REFERENCES

Burrough, P. A., and Frank, A.(1996). Geographic objects

with indeterminate boundaries, vol. 2. CRC Press.

Cresswell, T. (1996). In place-out of place: geography,

ideology, and transgression. U of Minnesota Press.

Deng, D. P., Chuang, T. R., Shao, K. T., Mai, G. S., Lin, T.

E., Lemmens, R., ... and Kraak, M. J. (2012). Using

social media for collaborative species identification

and occurrence: issues, methods, and tools. In

Proceedings of the 1st ACM SIGSPATIAL

International Workshop on Crowdsourced and

Volunteered Geographic Information (pp. 22-29).

ACM.

Erwig, M., and Schneider, M. (1997, January). Vague

regions. In Advances in Spatial Databases (pp. 298-

320). Springer Berlin Heidelberg.

Jones, C. B., Abdelmoty, A. I., Finch, D., Fu, G., and Vaid,

S. (2004). The SPIRIT spatial search engine:

Architecture, ontologies and spatial indexing. In

Geographic Information Science (pp. 125-139).

Springer Berlin Heidelberg.

Jones, C. B., and Purves, R. S. (2008). Geographic

information retrieval. International Journal of

Geographic information Science, 22(3), pp.219-228.

Jones, C. B., Purves, R., Ruas, A., Sanderson, M., Sester,

M., Van Kreveld, M., and Weibel, R. (2002). Spatial

information retrieval and geographical ontologies an

overview of the SPIRIT project. In Proceedings of the

25th annual international ACM SIGIR conference on

Research and development in information retrieval (pp.

387-388). ACM.

Lin, F. T., Liao, Y.P., and Lin, C.A. (2013) Using

Ontology in Planning Knowledge Management

System, the 5th Joint AESOP-ACSP Congress, Dublin,

Ireland.

Lynch, K. (1960). The image of the city. MIT press.

Ma, W. Y., and Chen, K. J. (2003). Introduction to CKIP

Chinese word segmentation system for the first

international Chinese Word Segmentation Bakeoff. In

Proceedings of the second SIGHAN workshop on

Chinese language processing, volume 17 (pp.168-171).

Association for Computational Linguistics.

Mata, F., and Claramunt, C. (2011). GeoST: geographic,

thematic and temporal information retrieval from

heterogeneous web data sources. In Web and Wireless

Geographic information Systems (pp. 5-20). Springer

Berlin Heidelberg.

Perry, M., Hakimpour, F., and Sheth, A., 2006. Analyzing

theme, space, and time: an ontology-based approach.

In Proceedings of the 14th annual ACM international

symposium on Advances in geographic information

systems (pp. 147-154). ACM.

Peuquet, D. J. (2001). Making space for time: Issues in

space-time data representation. In GeoInformatica,

5(1), pp.11-32.

Schneider, M. (2008). Fuzzy Spatial Data Types for

Spatial Uncertainty Management in Databases.

Handbook of research on fuzzy information processing

in databases, 2, pp.490-515.

Wang, F., and Hall, G. B. (1996). Fuzzy representation of

geographical boundaries in GIS. International Journal

of Geographic information Systems,10(5), pp. 573-590.

KEOD 2015 - 7th International Conference on Knowledge Engineering and Ontology Development

434