Semantic LBS: Ontological Approach for Enhancing Interoperability in ...

Semantic LBS: Ontological Approach for Enhancing Interoperability in Location Based Services Jong-Woo Kim1 , Ju-Yeon Kim1 , and Chang-Soo Kim2, 1

2

Interdisciplinary Program of Information Security, Pukyong National University, Korea {jwkim73, jykim}@pknu.ac.kr Dept. of Computer Science, Pukyong National University, Korea [email protected]

Abstract. Location Based services (LBS) is a recent concept that integrates a mobile device’s location with other information in order to provide added value to a user. Although Location Based Services provide users with much comfortable information, there are some complex issues. One of the most important issue is managing and sharing heterogeneous and numerous data in decentralized environments. The problem makes interoperability among LBS middleware, LBS contents providers, and LBS applications difficult. In this paper, we propose Semantic LBS Model as one of the solution to resolve the problem. Semantic LBS Model is a LBS middleware model that includes a data model for LBS POI information and its processing mechanism based on Semantic Web technologies. Semantic LBS Model provide rich expressiveness, interoperability, flexibility, and advanced POI retrieval services by associating POI Description Language (POIDL) ontology with heterogeneous domain specific ontologies.

1

Introduction

Recently, a lot of researches on the services for supporting the ubiquitous computing are undergoing in the various areas with the growing interest on ubiquitous computing. Context-Sensitive Computing is one of the key technologies supporting the ubiquitous computing which requires the information suitable for the user’s context, even in the restricted environment like mobile devices. Location Based Services especially provide the context sensitive information based on the user’s location. Location Based services (LBS) is a recent concept that integrates a mobile device’s location with other information in order to provide added value to a user [11,13]. Although Location Based Services provide users with much comfortable information, there are some complex issues. One of the most important issue is

Corresponding author.

R. Meersman, Z. Tari, P. Herrero et al. (Eds.): OTM Workshops 2006, LNCS 4277, pp. 792–801, 2006. c Springer-Verlag Berlin Heidelberg 2006

Semantic LBS: Ontological Approach for Enhancing Interoperability in LBS

793

managing and sharing heterogeneous and numerous data in decentralized environments. To resolve the problem, several efforts and studies have been made on improving efficiency of data management and establishing standards for information sharing in various operating environments each other [2,3,7]. In this paper we would like to propose Semantic Web approach to enhance interoperability through sharing LBS information. Semantic Web [1] is a technology to add well-defined meaning to information on the Web to enable computer as well as people to understand meaning of the documents easily. For our approach, we propose a Semantic LBS Model that is a LBS middleware model that includes a data model for LBS POI information and its processing mechanism based on Semantic Web technologies. We especially specify POI Description Language (POIDL) ontology that is a ontology-based description language. It can provide interoperability among LBS middleware, LBS contents providers, and LBS applications by allowing POI providers to describe their contents over domain specific ontologies. In section2, we introduce related work on LBS middleware for interoperability. In section 3, overview of Semantic LBS Model is described. In section 4, the description of LBS ontologies is given. In Section 5, we describe Semantic LBS middleware. Discussion and conclusion is given in Section 6.

2

Related Work

Location Based Services require integration of various technologies and standards. In order to make Location Based Services work, the industry had to overcome several challenges of both a technological and economic nature over the past years. Technologically, realizing LBS can be described by a three-tier communication model [13], including a positioning layer, a middleware layer, and an application layer. A middleware layer can significantly reduce the complexity of service integration because it is connected to the network and an operator’s service environment once and then mitigates and controls all location services added in the future. Moreover, a middleware layer can help LBS applications provide added value related to user’s location from heterogeneous data. As a result, it saves operators and third-party application providers time and cost for integrating application. Although most of the commercial LBS platforms have been implemented based on DBMS-based middleware, some approaches to efficiently manage heterogeneous data have been researched [2,3,7]. Although there have been a lot of studies on LBS middleware, LBS middleware has several challenges. First, providing users with added value to mere location information is a complex task, and the basic requirements of the variety LBS applications are numerous [11]. Second, it is difficult to manage LBS contents as a general data management in order to provide users with dynamic information frequently changed [12]. Next, it is difficult to share LBS information because the location-based services are operated in different processing methods, appropriative data exchange protocol, and various platforms [8].

794

3

J.-W. Kim, J.-Y. Kim, and C.-S. Kim

Semantic LBS Model

Semantic LBS is a platform that provides novel location-based services that provide not only LBS core services but also more enhanced POI(Point of Interest) retrieval service. A POI is a place, product, or service with a fixed position, typically identified by name rather than by address and characterized by type. A distinguishing feature of Semantic LBS is to retrieve the POIs with domain specific information by providing automatical interaction mechanism based on ontologies. Figure 1 shows the conceptual service model of Semantic LBS organized for the following three novel services. First, Semantic LBS provide the enhanced LBS Directory Service [10] that retrieves not only the POIs based on user’s location but also their domain specific information. It is difficult for current LBS to retrieve domain specific information because current LBS support searching only information stored as predefined data model. However, because Semantic LBS includes ontology-based data model that allow domain specific information to be specified, it is possible to retrieve domain specific information based on user’s complex requirements. Second, Semantic LBS is able to retrieve realtime updated information. In the Semantic LBS, contents providers publish the contents for each POI, and each POI is stored in decentralized system. The Semantic LBS provides the mechanism that enable agent to retrieve the contents stored in decentralized system and update information in POI repositories for LBS Directory Service [10].

Fig. 1. The conceptual service model of the Semantic LBS

The novel services of Semantic LBS are provided based on Semantic LBS Model. Semantic LBS Model is a LBS middleware model that includes a data model for LBS POI information and its processing mechanism based on Semantic Web technologies. For the data model, we specify two fundamental LBS ontologies for describing POI information: POIDL ontology and LBSTaxonomy ontology. POIDL is a OWL-based description language that allow POI providers to


795

describe their contents over domain specific ontologies. LBSTaxonomy is an ontology constructed as hierarchical taxonomy of POI types, and it is used to index instances described by the POIDL as POI types. POIs are actually described by the LBS ontologies, various domain specific ontologies, and association between them. Based on the data model, Semantic LBS Model provide some queries to retrieve and update POIs and their heterogeneous domain specific information.

4

Modeling Ontologies for LBS POIs

This section presents ontologies for describing LBS POIs (Points of Interest), and details some of the key classes and properties. LBS ontologies is composed of three kinds of ontologies: POIDL ontology, LBSTaxonomy ontology, and domain ontologies each with their unique XML namespace. POIDL ontology is most fundamental ontology in order to specify POIs, and it provides sufficient expressivness which can specify not only general information provided in current LBS model but also domain specific information. LBSTaxonomy ontology is an ontology which hierarchically classifies LBS POIs by types of services. It includes only necessary information for retrieving POIs based on classification and location. Domain ontologies are ontologies to specify domain specific information, and are used so that each contents provider specifies pertinent POI information conforming with its characteristics. 4.1

POIDL Ontology

POIDL is an ontology-based POI(Point of Interest) specification language that allows LBS contents providers to describe their information over domain specific ontologies. In addition, POIDL allow the information for query used to search and update domain specific information to be specified. This mechanism enable LBS to provide POI information changed frequently. POIDL ontology includes three classes: POI class, Content class, and ContentQuery class. POI class is used to describe general information about a POI and its domaindependent information. Figure 2 is an example of describing a POI using POI class. Instances of POI class specify POI information using five properties as follows. – poidl:name. This specifies a name of the POI that is specified in the instance of the POI class to identify the POI. In Figure 2, this property specifies that the instance describe the POI that has name of ’HotelA’. – poidl:isIncludeIn. : This property specifies a type of the POI instance. This associates the POI instance with a class of LBSTaxonomy ontology described in Section 4.2. The LBSTaxonomy ontology is an ontology that specifies hierarchical classification of POI types. In Figure 2, this property specifies that the type of the POI instance is ’Hotel’. – poidl:hasContent. This property is used to associate an instance of the Content class that specifies domain specific information about the POI, and a POI can have more than one contents. In Figure 2, this property associates

796


Fig. 2. An example instance of POI class

two contents, ’Grade HotelA’ and ”EmptRoom HotelA”, that are instances of Content class. – The four location properties - poidl:rectLeft, poidl:rectRight, poidli:rectTop, and poidl:rectBottom describe the boundary of a POI. Location can be described by various form, but we suppose that location is represented by rectangular coordinates used in geographical map. In order to indicate geographical location through various coordinates systems and address, we can specify an ontology for location and associate an instance of location ontology. Content class is used to describe a domain specific information for a POI, and an instance of the Content class specifies type and value of the POI. The Content class provides rich expressiveness and flexibility because instances of the class specify domain specific information through referring domain specific ontologies and its instances. Also the Content class associates an instance of the ContentQuery class that specifies a query for retrieving current value in its domain. Figure 3 is two example instances of the Content class, which describe grade of ’HotelA’ and the number of empty room of ’HotelA’ respectively. the grade is static information, while the number of empty room is a dynamic information that changes frequently. The three properties for describing an instance of the Content class is as follows. – poidl:typeofcontent. This property specifies a type of domain specific information and associates a class of the domain specific ontology of the POI. In Figure 3, this property describe that an instance of the Content class, ’Grade HotelA’ specifies the grade of ’HotelA’. – poidl:valueofconent. This specifies a value of domain specific information of the POI. In example of Figure 3, This property specifies that the grade of ’HotelA’ is ’Grade 1’. And we can know that the number of empty rooms has to be retrieved because it is dynamic information that changes frequently. – poidl:query. This associates an instance of ConentQuery class for querying domain specific information of the POI. For example, in order to retrieve


797

Fig. 3. An example instance of Content class

Fig. 4. An example instance of ContentQuery class

the number of empty rooms for ’HotelA’, an instance of ContentQuery class, ’Query EmptyRoom HotelA’, can be referred. The ContentQuery class is used to describe a query for retrieving domain specific information from its domain. This class enable domain specific information to be retrieved and automatically updated. The query language to retrieve information from RDF documents such as RDQL is represented as triple, same as RDF. The ContentQuery class specifies a triple for a query - subject, predicate, and object. Figure 4 is an example instance of the ContentQuery class. The three properties for describing an instance of the ContentQuery class is as follows. – poidl:subject. This property describes a subject of triple for query. – poidl:predicate. This describes a predicate of triple for query. – poidl:object. This property describes an object of triple for query. 4.2

LBSTaxonomy Ontology

The LBSTaxomomy ontology is used to guide the taxonomy of POI entities in the LBS domain. An instance of the LBSTaxonomy ontology specifies the information about the POI used in Directory Service of Semantic LBS. An instance of the LBSTaxonomy ontology includes the basic information of the POI such as

798


Fig. 5. An Example instance of LBSTaxonomy ontology

name, and location information to efficiently retrieve POIs as taxonomy. It also specifies the information to refer the instance of the POIDL ontology that specifies domain specific information of the POI and association information used to specify general retrieval patterns of user. In this work, the LBSTaxonomy ontology refers to the hierarchical classification code defined by National Geographic Information Institute (NGI) in Republic of Korea, and we extend it. For example, Hotel class is a subclass of ServiceFacility class, and ServiceFacility is a subclass of Structure. Figure 5 shows an example instance of the LBSTaxonomy. LBSTaxonomy ontology includes seven properties as follows. – tax:details. This property associates an instance of the POI class in POIDL ontology. This property provide connectivity with detailed information for POIs. – tax:name. This describes identification of a POI. The object of this property is same as the object of ’name’ property of POIDL ontology. For example, the ’HotelA’ in Figure 5 indicates the identification of a POI. – The four location properties - tax:rectLeft, tax:rectRight, tax:rectTop, and tax:rectBottom describe the boundary of a POI. The object of this property is same as the object of location properties of POIDL ontology. LBSTaxonomy ontology is specified to be basically used in Semantic LBS Model, but applications can specify application-specific ontologies as their purposes and features. 4.3

Domain Specific Ontology

The domain specific ontologies define concepts for each domain and relationship between them. The domain specific ontologies can be defined by contents provides or the Standard Organizations. Because LBS contents is different as their domain, the domain specific ontologies help domain specific information to be represented more exactly. The domain specific ontologies is referred by the instances of POIDL ontology to specify the domain specific information.


5

799

Semantic LBS Directory Service with LBS Ontologies

We approach the enhancement of LBS Directory Service by building the middleware that provides retrieval services for Semantic LBS data model as shown in Figure 6. Semantic LBS Model provides not only general POI retrieval queries but also some advanced queries based on expressiveness and interoperability of Semantic LBS data model. Semantic LBS Model enhances general queries of the Directory Service. The Directory Service of location-based services provides a search capacity for one or more Points of Interest (POI), and it provides several kinds of queries as range of retrieving POIs. Semantic LBS Model enable to retrieve POIs and their domain specific information with more complex conditions, while current LBS Models provide a simple search capacity that can retrieve POIs based on only location. The Semantic LBS Model also provides a contents query that retrieves more detailed and domain specific information about a POI. The query enable to acquire dynamic information changed frequently and automatically update POI directories.

Fig. 6. The architecture of Semantic LBS Middleware

The query modules in Figure 6 are implemented using Jena, Joseki and RDQL templates. Jena API is a Java application programming interface that creates and manipulates RDF documents, and Joseki is a Java client and a server that implements the Jena network API over HTTP [4,9]. We can semantically search the instances of RDF documents through RDQL, a Qurey Language for RDF, which is Jena’s query language [4]. Figure 7 shows a simple application of Semantic LBS, Hotel Finder. The application provides hotel information including domain specific information such as room type, meal type, and price (Figure 7 (b). It also provides the map service that utilize Mobile GIS module developed in our previous work (Figure 7 (c)) [5,6].

800


Fig. 7. An example application of Semantic LBS

6

Discussion and Conclusion

Mobile users expect that Location Based Services provide more specific information exactly. Therefore, LBS middleware models have to provide domain specific information even if LBS data is heterogeneous and numerous. Semantic LBS Model is a LBS middleware model that includes a data model for LBS POI information and its processing mechanism based on Semantic Web technologies. For the data model, we specified two fundamental LBS ontologies for describing POI information: POIDL ontology and LBSTaxonomy ontology. POIDL is a OWL-based description language that allow POI providers to describe their contents over domain specific ontologies. LBSTaxonomy is an ontology constructed as hierarchical taxonomy of POI types, and it is used to index instances described by the POIDL as POI types. POIs are actually described by the LBS ontologies and heterogeneous domain specific ontologies. Based on the data model, Semantic LBS Model provide some queries to retrieve and update POIs and their domain specific information. Main contributions of our approach include: – Expressiveness of POIs: Semantic LBS Model provides sufficient expressiveness. The POIDL in Semantic LBS Model provide more expressive and more flexible description mechanism for POI information that enable domain specific information to be described. – Interoperability: Semantic LBS model supports interoperability through information sharing in decentralized environments. Semantic LBS Model uses HTTP that is standard data exchange protocol of Web and shares LBS information through URI and ontologies. – Benefits of POI retrieval: LBS Model provides not only basic queries for retrieving POIs but also some advanced mechanism that retrieves POI information: retrieving domain specific information even if the information changes frequently, and automatically updating domain specific information of POIs for LBS directory service. The ontology-based data model enable Semantic LBS to provide the advanced functions.


801

– Flexibility: Semantic LBS allow the domain specific ontologies to extend without modifying the middleware and applications. Moreover, because POIDL specifies templates for retrieving domain specific information of each POI, Semantic LBS is able to appends POI information easily. Acknowledgement. This work was supported by the Korea Research Foundation Grant funded by the Korean Government(MOEHRD) (KRF-2005-908D00057).

References 1. Berners-Lee, T., and et. al.: The Semantic Web, Scientific American (2001) 2. I. Burcea and H.-A. Jacobsen: L-ToPSS: Push-Oriented Location-Based Services, In Proc. of the 4th VLDB Workshop on Technologies for E-Services (TES’03), Lecture Notes in Computer Science, No. 2819, Springer-Verlag (2003) 3. N. Davies, S.P. Wade, A. Friday, and G. S. Blair: Limbo: A Tuple Space Based Platform for Adaptive Mobile Applications, In Proceedings of the International Conference on Open Distributed Processing/Distributed Platforms (ICODP/ICDP) (1997) 4. Jeremy J. and et. al.: Jena: Implementing the Semantic Web Recommendation, ACM (2004) 5. J.W. Kim, et. al.: The Efficient Web-Based Mobile GIS Service System through Reduction of Digital Map. Lecture Notes in Computer Science, Vol. 3043, SpringerVerlag (2004) 6. J.W. Kim, et. al.: Location-Based Tour Guide System Using Mobile GIS and Web Crawling, Lecture Notes in Computer Science, Vol. 3428, Springer-Verlag (2005) 7. H. Leung, I. Burcea, and H.-A. Jacobsen; Modeling Location-based Services With Subject Spaces, In Proceedings ofCASCON Conference, IBM Publisher (2003) 8. Mabrouk, M. et. al.: OpenGIS Location Services (OpenLS): Core Services, OpenGIS Implementation Specification, OGC (2004) 9. McBride, B.: Jena: A Semantic Web Toolkit. IEEE INTERNET COMPUTING. IEEE (2002) 10. OpenLS Initiative.: A Request for Technology in Support of an Open Location Services Testbed. OGC (2000) 11. J. Schiller: Location-Based Services, Elsevier, p.1-5 (2005) 12. Shiowyang Wu, Kun-Ta Wu.: Dynamic Data Management for Location Based Services in Mobile Environments. Proc. 2003 Int. Conf. Database Engineering and Applications symposium. IEEE (2003) 13. S. Spiekermann: General Aspects of Location-Based Services, Location-Based Services, Elsevier 2005.