Semantic Retrieval of Geospatial Datasets Based on a Semantic

Semantic Retrieval of Geospatial Datasets Based on a Semantic Repository Julio Vizcarra, Miguel Torres, Rolando Quintero Intelligent Processing of Geospatial Information Lab-Centre for Computing Research-National Polytechnic Institute, Mexico City, Mexico [email protected],{mtorres, quintero}@cic.ipn.mx http://www.cic.ipn.mx

Abstract. In this paper, we propose a methodology to semantically search geospatial datasets by means of their metadata. It consists of structuring a semantic repository in order to provide the inclusion mechanisms of distributed data to be retrieved, as well as the extraction of those geographic objects with respect to their conceptual similarity. The approach proposes a conceptual measure called DIS-C algorithm to determine the conceptual similarity among concepts. The features of geographic objects are conceptualized in ontology on the domains: thematic, spatial and temporal. The ontology is populated by using FGDC specification with instances of geographic objects distributed in several locations on a network. In the retrieval process according to a query, the objects are presented as a ranking list, which provides other results close semantically by means of radius of search and it avoids empty results. The approach has been implemented in a web system called SemGsearch.

1

Introduction

Nowadays, there are a large volume of geographic data collected, which are obtained by different technologies such as GPS (Global Positioning System), satellite images, geographic databases, maps in analog format, among other sources [18], and not only by new spatial information systems, but also data collection technologies that are becoming more sophisticated. In addition, geospatial data are an important issue of any decision support system (DSS), they can be considered as crucial features for planning and decision-making in a variety of applications. It is important to mention that the development of technologies for integration, tools for management and analysis are increased in recent years [10]. On the other hand, there are not agreements in the representation of geospatial semantics in maps and generally in geospatial representations. For example, there are different organizations with an accuracy to draw certain lines, points or polygons on a plane, in order to represent cities, water supply wells, or altimetry points in the network, transmission lines, road infrastructure, and so on [15]. However, it does not exist consensus or agreement between organizations or groups of experts about the

50 J. Vizcarra et al. ____________________________________________________________________________

meaning, semantics or ontology of these conceptual representations, which cause that each organization has a particular conceptualization. For instance, at the same time many organizations handle the concept "artificial lakes" and others the concept "dam". The problem now is not how to accurately represent a geographic feature, the problem is when two spatial representations or geographic databases representing the same, or have a common semantic unit, or the user is familiar with a naming (semantic annotation) and he understands other cartographic information made by another organization or another user. This semantic union is the basis for obtaining true interoperability and exchange of geospatial data among different users. Therefore, geographic information systems (GIS) are not exempt from this problem in order to manage and process these data [23]. Most GIS are not originally designed to work with semantic processing, there are some problems of interoperability, while the integration of heterogeneous geospatial data sources can not be achieved in these systems [5]. This is mainly because each GIS provides specific requirements for the representation of its data, such as its format and specific query language [3]. The problem in information retrieval is that users can be easily overwhelmed by the amount of information available. The processing time of irrelevant information in documents such as geospatial databases, geospatial images, text and tables retrieved by an information retrieval system is very chaotic and this condition is the result of inaccuracies in the representation of the documents in the repository as well as confusion and imprecision in user queries, since users are frequently unable to express their needs efficiently. These facts contribute to the loss of information and the provision of irrelevant information. According to the problems mentioned above, this paper proposes a methodology for semantic search geospatial data by means of their metadata. The metadata are structured into a semantic repository, adding a priori knowledge of the thematic, spatial and temporal domains in order to provide mechanisms that include distributed data to be retrieved and the extraction of these geographic objects with respect to their conceptual similarity from several servers distributed. The paper is organized as follows: Section 2 presents the state of the art related to the work in this field. Section 3 describes the proposed methodology to semantically retrieved geospatial information. In Section 4, the experimental results are depicted. Finally, the conclusion and future works are outlined in Section 5.

2

Related Work

In this section several approaches that are used for information retrieval in different data sources to solve the problem of semantic heterogeneity have been analyzed. Some of these works have exploited the metadata, designing descriptors of the information contained in the repositories. Other projects have proposed the use of ontologies, but the common characteristic is the intelligent search, so that models have been developed to retrieve an object or concept from various sources.

Semantic Retrieval of Geospatial Datasets Based on a Semantic Repository 51 ____________________________________________________________________________

In [12] a theory of similarities related in the context on modeling web services is presented. The implementation of the theory is performed using the framework for WMSL, which is based on the intersection of logical descriptors. For the process to give the measure of similarity of a concept, it extracts the attributes or properties of its specification in the WMSL [18]. Later, a matrix is formed in order to compare with other concepts and their respective matrix. For the identical pairs, it is necessary to assign a weight, similar or different to each pair. In [16] a framework of semantic annotations is proposed. Thus, in [13] an approach to integrate and retrieve information on distributed heterogeneous resources on the Web is described. The semantic annotations defined by the knowledge base within the knowledge discovery processes and control over it are called entities, which can be described generally interconnected to define a process of discovery of pieces of knowledge in data with little or no human assistant. Another case is presented in [4], which studied the design, syntax and implementation of the semantics and integration within e-business [2]. The integration is performed reconstructing the notion of object-relation with shared components. There are high-level languages to describe knowledge and integrate it into the semantic web by means of techniques of positional and slotted and Artificial Intelligence. A language that defines the concepts and semantic relations between them in order to integrate accurate data between different sources is created. On the other hand, the work presented in [19] is focused on the discovery and retrieval of spatial data in distributed environments on Spatial Data Infrastructures (SDI). The discovery and recovery tasks are based on three steps: in the first step the user's search terms are mapped to concepts in domain ontology based on the hybrid ontology approach [14]. In the second step, the concepts are extended on the basis of the hierarchy of concepts in the domain ontology, and the third step consists of the expansion of the query, adequate descriptions of geographic information are sought and returned to the users. If the results are adequate, the search is over.

3

The SemGsearch Methodology

To achieve a semantic search, it is necessary to process geospatial data sets at a conceptual level; nevertheless, a geographical object can be described in many ways, regarding the degree of knowledge, abstraction and interpretation desired. The conceptualization of a domain is used to integrate information. In the geographic domain, we propose to apply the GEONTO-MET approach [20] in order to build geographic ontologies. Additionally, we proposed the DIS-C (Conceptual Distance) algorithm to compute the distance between concepts defined in the retrieval process, when the user performs a query. This algorithm avoids obtaining empty answers, whether the exact concept is not found in the sources. Therefore, users will obtain as response a concept that can be conceptually closer. The conceptual structure and stages of SemGsearch approach is shown in Fig. 1.

52 J. Vizcarra V et al. ___________________________________ _____________ ___________________________________

Fig. 1. Concepptual structure of o SemGsearch..

Conceeptualization.. This stage coonsists of buillding the know wledge base, represented direectly by ontollogy. This connceptual structure stores infformation relaated to the geograaphic domain. It also compuutes the conceeptual distancee table as a baasis for the custom m query. Synthesis. This stagge involves the instantiation n of objects thhat represents geographical cooncepts of thee domain by using u the metaadata that desscribe each daata source. In other words, thee stage has thhe function off populating the t ontology with geographiic objects locaated within seeveral data so ources that com mpose the sem mantic repositorry. Analyysis. At this staage the reposiitory is used fo or semanticallly searching ggeographic objectss on domains,, using the conncepts related to the query request. r The coonceptual fram mework withh the processees of the threee stages is depicted in Fig. 2. Thhe conceptuaalization stagee uses the app plication ontoology that contains the knowledgge base of thhe domains deescribed in th he conceptuallization. By aapplying a conceptuaal similarity algorithm, a we build a graph h that determiines a distancce value to each concceptual relatioonship from thhe ontology built b by GEON NTO-MET. Finally, the FloydWaarshall algorithhm, describedd in [9] is useed to determinne the smallesst value or cost of eaach concept or o node to anoother, indicatin ng a similarityy value. This algorithm reduces the t performannce with largee number of nodes; n due to it works withh matrices. The synthesis stage picks the geosspatial data th hrough its Feederal Geograaphic Data Committeee (FGDC) [88]. This speciification defin nes by means of a metadatta description, the main featuress of a geograaphic object to o be retrievedd, and it is ussed in the populatioon task for reading the meetadata file, which w describees a geographhic object. Later, ann instance for each geograpphic object iss created, afteer this processs each of these objects is linkedd to the other three domain ns: thematic, spatial s and tem mporal by means off their metadatta keywords in i order to sem mantically rettrieve geospattial data in the next stage. s In the analysis stagge the searchiing is perform med by the decomposition d n from the query to concepts. Addditionally, thee vector creatted by the conncepts will bee used for searchingg concepts witthin the ontollogy together with the geoggraphic objectt linked in the synthhesis stage. Moreover, M the retrieval r proceess is based on o the expanssion of the conceptuaal distance am mong conceptss, increasing the t radius. Thhe first approxximation is

Semantic Retrieval R of Geoospatial Dataseets Based on a Semantic S Repossitory 53 ___________________________________ _____________ ___________________________________

to find eqqual or exact concepts for the t conceptuaal distance k = 0. Later, eacch domain k > 0 from m the resultingg FloydWarshhall table in th he conceptualization is extennded.

F 2. Concepttual framework of SemGsearchh. Fig.

3.1

Coonceptualizattion Stage

In this staage, the domaains that definne geographic objects are coonceptualized. It is possible to use u a variable number of doomains; howeever, in this work w we only uused three domains: thematic, sppatial and teemporal, whiich were connceptualized using the GEONTO O-MET methoodology. It is important to mention wheether the numbber of domains is increased, thee semantic grranularity of th he ontology is i improved aand the retrieval prrocess is enricched. This factt allows refiniing a query annd making it eeven more specializeed. 3.1.1. DIS-C: Conceptual similaritty algorithm The concceptual distancce is defined by the space that separatess two conceptts within a specific conceptualizat c tion [1]. The main idea of the algorithm m is to establissh the distance value of each relationship witthin the ontollogy, and trannslate it into a weighted directed graph, g where each concept becomes a no ode and each relationship bbecomes a pair of eddges. The cllassical step for f computingg the minimu um distances is to find thee distance among cooncepts that arre not directlyy related. Thus, the steps to compute the conceptual distancce are the folloows: 1. For eaach type of reelationship, it is necessary to assign a conceptual c distance for each relationship r w within the conceptualization n. For examplle, if it has thhe relation "is" annd "road is a communicattion route", we w can set thee distance of "road" to "comm munication rouute" and “com mmunication route" to "road". As an illlustration, we can establiish that distance(road d d, communication_route)=1 and r Forrmally, K (C , , R ) is a trriplet that distancce(communicaation_route, road)=0. defines a conceptuualization, whhere C is thee set of conccepts, is tthe set of

54 J. Vizcarra et al. ____________________________________________________________________________

relationships types and R is the set of relations in the conceptualization. For each , it has to set the values of to the relationship in type of relationship the normal sense, and

to the relationship

in reverse sense. In the example

is

is

above =0 y = 1. 2. The directed graph is created GK (V , A) for the conceptualization K. First, it is added to each item c C as a vertex in the graph GK , i.e., V = C . Now, for each relationship a b R , where a, b C , edges (a, b, ) and (b, a, ) are added. 3. Once the graph has been built, we apply the FloydWarshall algorithm to process the minimal distance between two nodes. The table of minimum distances between each pair of vertices, which are directly mapped to the concepts in the ontology is created. So the result is the conceptual distance disseminated to all concepts in the conceptualization K. 3.1.2. Application of conceptual similarity in GEONTO-MET GEONTO-MET [19] is composed of three axiomatic relations: “is”, “has” and "does". The relation “is”, as we mentioned in the example “road is a communication route”, it establishes the distance of “road” to “communication route” and “communication route” to “road”. It is defined by distance(road, communication_route)=1 and distance(communication_route, road)=0. Thus, we propose that if a(is )b RR then is

is

(a, b) = 1 . For the relation “has”, it defines properties, in which (a, b) = 0 and the distance is inversely proportional to the number of occurrences of the concept that “has” presents. For example, if an urban area “has” a street of the first order, then the conceptual distance between the concept “urban area” and the term “street” will be inversely proportional to the number of streets that the urban area presents. That is, if 1 a ( has )b R R , then has (a, b) = , where R ( p ) is the occurrences number of R( p) the property p = a ( has )b in RR (this value is normally of 1). On the other hand, the conceptual distance of “street” to “urban area” will also be inversely proportional to the number of “streets”, Otherwise, “urban area” is directly proportional to the number of properties that “urban area” contains (streets, buildings, parks, etc.) If card ( P(a)) has ( a, b) = a ( has )b RR , then , where P(a) = x | a( has ) x RR for R( p) any concept x C and R ( p ) is the occurrences number of the property p = a ( has )b in RR . The relation “does” defines abilities and the conceptual distance is computed in both senses of the relationship, inversely proportional to the number of times that one abilities is referred by one concept, likewise, in the inverse relationship, it is directly proportional to the total number of abilities that a concept contains. For instance, we consider the ontology depicted in Fig. 3, in which the blue arrows indicate the relation “is” and the red arrows indicate the relationship “has”. The con-

Semantic Retrieval R of Geoospatial Dataseets Based on a Semantic S Repossitory 55 ___________________________________ _____________ ___________________________________

ceptual distance d is defi fined between the pairs of concepts c by thhe DIS-C algoorithm and as a resullt a weighted directed graph is created (see Fig. 4). Once O the graphh is generated, it is necessary to compute the shortest s distan nce C (a, b) .

Fig. 3. Geosspatial ontologyy for the divisio on of a country e.g. Mexico.

F 4. Weightedd directed graphh obtained from Fig. m applying the DIS-C D algorithm m.

Once the graph with thhe weights corrresponding to o the conceptuual distance iss obtained, the next step is to traansform it in a matrix. Laater, it is necccesary to meeasure the axiomaticc relations. To illustrate thhe procedure, we have in the example above the (concept 1, concept 2) 2 related unnder the relattion “is”, acccording to D DIS-C, the “concept 1” to “conceppt 2”, the concceptual distancce or weight w = 1 and 0 coonversely. In the case of (conccept 2, conceppt 4) related under u the relattion “has” thee “concept 1 1” to “conncept 2”, the conceptual c disstances w is th he weight , that is,, it has 1/1 R(aa b) card ( P(a)) and convversely , is 1//1. Finally, th he concepts (concept ( 2, cconcept 3) R(aa b) 1 under thee relation “does” the weighht w of “concept 1” to “conncept 2” is , it R ( a b)

56 J. Vizcarra et al. ____________________________________________________________________________

card ( P(a)) is 2/1. The next step is to determine the lowest R ( a b) weight w between two concepts a and b belonging to a given domain ontology. The process is carried out using the FloydWarshall algorithm, which considers as input the resulting graph of the algorithm as a DIS-C matrix between concept weights w, Cx, Cy another belonging to the same domain indicating the shortest path or minimum weight between any pair of nodes Cx, Cy that belong to the graph DIS-C. The matrix that indicates the minimum weight of a concept Cx to a concept Cy is described in Table 1. In addition, we present the DIS-C algorithm in Table 2. has 1/1 and conversely

Table 1. Matrix generated by FloydWarshall algorithm using the DIS-C graph, indicating the shortest path.

Concepts C1 C1 0 C2 1 C3 3 C4 3 C5 2

C2 0 0 2 2 1

C3 1 1 0 3 2

C4 1 1 3 0 2

C5 1 1 3 3 0

Table 2. The DIS-C algorithm. 1.

GO

weighted directed graph

2.

RR

relations in O

3. 4.

For all a b c RR do If =is then

5.

Add (a, b,0) to GO

6.

Add (b, a,1) to GO

7. 8. 9.

3.2

Else if

= has then

1 ) to GO R ( a b) card ( P (a )) ) to GO Add (b, a, R ( a b)

10.

Else if

= does then

11.

Add (a, b,

1 ) to GO R ( a b)

12.

Add (b, a,

card ( H (a)) ) to GO R ( a b)

End if 13. 14. End for 15. Return Shortest path ( Go )

Add (a, b,

Stage of Synthesis

This stage generates instances from the concepts described in the ontology (previous stage). The task is performed for each site or server where they are data sources. It is described as follows: 1. For each data source located in the repository, the description of geographic objects is obtained using the metadata file that accompanies each geographic object and it is mapped on FGDC domain (see Fig. 5).

Semantic Retrieval R of Geoospatial Dataseets Based on a Semantic S Repossitory 57 ___________________________________ _____________ ___________________________________

For eaach domain of metadata, thhe set of keyw words is idenntified. These keywords are maapped with concepts in the others o domain ns (see Fig. 6)).

Fig. 5. Insstances of geospatial data in FGDC F Fig. 6. Geographic G objeects linked in thhe ontology. domain.

3.3

Sttage of Analyysis

In the annalysis stage, the t tasks of semantic searcching and retrrieval as well as the results dispplaying are carrried out. The stage is comp posed of five basic b processees: 1. The cooncepts and their t domains are obtained from the queery, after thatt, a vector that deefines the geoggraphic objectts is created. 2. With the t resultant vector v previouusly, the next step is to determine each cconcept to its resppective domaiin. 3. Whethher the conceppt in the queryy has an exact match with a concept in thhe domain, this is the closest coonceptual disttance, i.e., k = 0. The proceess continues extending for k > = 0 (see Fig. 7). 4. When the radius aree increased to values k > = 0 , more conccepts are obtaained, thus severaal instances of o geospatial objects in th he data sourcces are retrieeved. This proceddure is appliedd to each dom main, then a list l of conceptts that are sem mantically relatedd is created byy means of thee intersection among a domainns (see Fig. 8)). 5. The laast task of thhe searching is i to return the t entire set of instances from the previoous procedures for k = 0 thhat representss an exact maatch and afterr that, for valuess k > 0 that aree semanticallyy close (see Fig g. 9).

58 J. Vizcarra V et al. ___________________________________ _____________ ___________________________________

Fig. 7. 7 Conceptual diistance extendinng the searchinng range

Fig g. 8. Semantic inntersection amoong the sets of geogrraphic conceptss

F 9. Instancees retrieved from Fig. m the repositoryy.

4

R Results

As case study, s the them matic domain is composed of o urban areass, communicaation infrastructure,, and forest reesources usingg the specificaation from INE EGI (Mexicann National Institute of o Statistics, Geography G andd Informatics)) [11]. The spatial domain is referred to Mexicco with differrent classificaations: regionaal division byy CONABIO (Mexican National Commission for Knowledgge and Use of Biodiversity) [7], spatial diistribution area by CFE C (Mexicann Federal Elecctricity Comm mission) [6] and economical area as in the classiification from m INEGI. Finaally, the tempo oral domain iss conceptualizzed taking into accouunt the Gregoorian calendar.. The seemantic reposiitory is structuured on SemG Gsearch by meeans of an adm ministrator interface,, which adds servers, loadss domains thaat are built unnder GEONTO O-MET in OWL forrmat [17]. Thee system allow ws the possibiility of queryinng the metadaata files of the geogrraphic objectss located in thhe servers. The SemGsearrch administraator is depicted in Fig. 10.

Semantic Retrieval R of Geoospatial Dataseets Based on a Semantic S Repossitory 59 ___________________________________ _____________ ___________________________________

F Fig.10. SemGssearch adminiistration interfface. As an example the next n query rettrieves the geo ospatial objecct "boulevard"" in the location off "Veracruz staate, Mexico" (see ( Fig. 11). Due to “boulevard” at “Veeracruz” is not storedd in any serveer; this query will w not retriev ve any instancce with an exaact matching (k = 0). 0 In this casee, to avoid em mpty answers, the conceptuaal distance is applied in order to retrieve r similaar concepts acccording to thee conceptualizzation.

Fig. 11. Semantic searchh: thematic "bouulevard", spatiall "Veracruz, Meexico", year "20006"; k = 0

Now, for a value k = 1, the semaantic search iss increased in the partition of the ontology, acccording to thhe conceptualiization, within n spatial domaain “Eco-regioonal Division”. Thhere is a simiilarity under the t classificattion of “Warm m Humid Junngles”, the concepts "Chiapas statte" and "Veraccruz State" are the closest states s semantiically with respect too the others (see Fig. 12).

Fig. 12. Ontology partittion of the classs “Eco-regionall Division” and “Warm Humidd Jungles”

60 J. Vizcarra V et al. ___________________________________ _____________ ___________________________________

Accordding to the increment of thhe conceptual distance for k > 0, the nextt search is made as ”Boulevard” that is located in “Veracru uz”, this geosspatial object is already located inn one server in i the repositoory. The resullts that accom mplish the requuest query by meanss of semantic similarity aree depicted in fig. 13 and fig. f 14. In thiss case, for "boulevarrd" in "Veracrruz", with k = 1 and k = 2 reespectively.

Fig. 13. Seemantic search: "boulevard", "V Vera acruz", "2006" with w k=1

Fig. 14. Semantic seearch: "boulevaard", "Veracruz", "20006" with k = 2

In thee retrieval proocess the geoographic objeects can be downloaded d ddirectly or viewed byy means of a snapshot s suchh as is depicted d in the Fig. 15.

Fig. 15. Snappshot of a geog graphic object.

5

Conclusions

In this paaper, a methoddology focusedd on the taskss of semantic integration i and retrieval of heterogeneous data sources, whicch can be locaated on differeent servers is proposed. Similarlyy, the main gooal of this papper is to be able for sharingg geospatial innformation by meanss of semantic retrieval mecchanism, using g a conceptuaal representatioon. In this approachh, a mechanism m to measure the conceptuaal distance callled DIS-C alggorithm is proposedd. This algorithhm has the fuunction of meaasuring how close c two conccepts conceptuallyy are. It allowss the system to resolve queries accordingg to k values tthat represent a connceptual distannce. This metthod avoids reeturning emptyy requests to tthe users.

Semantic Retrieval of Geospatial Datasets Based on a Semantic Repository 61 ____________________________________________________________________________

The present work leads to an easier and faster way of geographic objects searching, it aids to solve multiple conceptualizations integrating different domains. In this proposal, it is not important to know the expertise degree of the users, because the approach is in charged of integrating and conceptualizing the domains by means of describing the metadata. As case study, we implemented the SemGsearch system for testing the proposed approach. This application provides ranking list of the geographic objects semantically retrieved. Future work is oriented towards applying this methodology in particular cases of the semantic web. The conceptual distance is an important issue that covers semantic information integration and semantic interoperability among applications and information in the web.

6

Acknowledgments

This work was partially sponsored by the National Polytechnic Institute (IPN), the National Council for Science and Technology (CONACyT) under grant 106692, and Research and Postgraduate Secretary (SIP) under grants 20120563 and 20121661.

References 1. Abu-Hanna A, Jansweijer WNH (1994) Modeling domain knowledge using explicit conceptualization. IEEE Expert, 9(2): 53-64 2. Amor D (2000) The e-business (r) evolution: living and working in an interconnected world. Prentice Hall PTR Upper Saddle River, NJ 3. Bishr Y (1998) Overcoming the semantic and other barriers to GIS interoperability. International Journal of Geographical Information Science, 12(4): 299-314 4. Boley H, et al (2010) Integrating positional and slotted knowledge on the semantic web. Journal of Emerging Technologies in Web Intelligence, 2(4): 343-353 5. Buccella A et al. (2007) Integration of Geographic Information Systems. In IX Workshop of Researchers in Computer Science, pp. 397 6. CFE (Mexican Federal Electricity Commission), http://www.cfe.gob.mx/. Accessed 20 June 2012 7. CONABIO (Mexican National Commission for Knowledge and Use of Biodiversity), http://www.conabio.gob.mx/. Accessed 20 June 2012 8. FGDC (Federal Geographic Data Committee), http://www.fgdc.gov/. Accessed 20 June 2012 9. Floyd RW (1962) Algorithm 97: shortest path. Communications of the ACM, 5(6): 345 10. Heywood I et al. (2006) Geographical information systems. Pearson Education Limited 11. INEGI (Mexican National Institute of Statistics, Geography and Informatics) (2012). http://www.inegi.org.mx/. Accessed 20 June 2012 12. Janowicz K (2007) Similarity-based retrieval for geospatial semantic web services specified using the web service modeling language (wsml-core). The Geospatial Web-How Geo-Browsers, Social Software and the Web, 2: 235-246 13. Kiryakov A et al. (2004) Semantic annotation, indexing, and retrieval. Web Semantics: Science, Services and Agents on the World Wide Web, 2(1): 49-79

62 J. Vizcarra et al. ____________________________________________________________________________ 14. Liang Y et al. (2007) Hybrid Ontology Integration for Distributed System. In Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing. SNPD 2007, 1(1): 309-314 15. MacEachren AM (2004) How maps work: representation, visualization, and design. The Guilford Press 16. Malik SK et al. (2010) Semantic Annotation Framework For Intelligent Information Retrieval Using KIM Architecture. International Journal of Web & Semantic Technology (IJWest), 1(4): 12-26 17. McGuinness et al. (2004) OWL Web Ontology Language Overview. W3C recommendation, 2004-03 (10), http://www.w3.org/TR/owl-features/. Accessed 20 June 2012 18. Sabbouh M et al. (2007) Web Mashup Scripting Language. In Proceedings of the 16th International Conference on World Wide Web, ACM, pp. 1305-1306 19. Santos JM et al. (2003) Fuentes, tratamiento y representación de la información geográfica, “Sources, treatment and representation of the geographic information”.Universidad de educación a distancia UNED, Madrid pp. 421 20. Torres M et al. (2011) GEONTO-MET: an approach to conceptualizing the geographic domain. International Journal of Geographical Information Science, 25(10): 1633-1657 21. Vilches LM et al. (2009) Towards a semantic harmonization of geographic information. Treballs de la Societat Catalana de Geografía, pp. 727-736 22. Zhan Q, Zhang X, Lic D (2008) Ontology-Based Semantic Description Model for Discovery and Retrieval of Geospatial Information. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 32