| GISdevelopment.net ---> Technology ---> Geographic Information System |
Semantic Interoperability of Geographic Information
Sumit Sen
Institute for Geoinformatics (IFGI), Robert-Koch-Str. 26-28, 48149 Münster, Germany
Email: sumitsen@uni-muenster.de
Sumit Sen
Institute for Geoinformatics (IFGI), Robert-Koch-Str. 26-28, 48149 Münster, Germany
Email: sumitsen@uni-muenster.de
Geographic Information (GI) is characterized by cross domain applicability and the eco-nomic value of such information is linked to the wider usage of such information. With increas-ing efforts to provide GI data and processing through web based services, it has become essen-tial to have a closer look at the term interoperability, its various forms and the notion of seman-tic interoperability in particular. Recent research focus on these issues has helped to outline some major challenges of semantic interoperability of GI and approaches to solve these.
1. GI usage and Information communities
Users of GI come from different domains and are ever expanding. Traditionally such in-formation resided locally with the users and human interfacing provided the means to usage across domains. Such user domains could be at various levels ranging from industry verticals, cultural-linguistic communities or even departments in the same organization. Such domains can be characterized by the term information communities [1].
Access and usage of GI has undergone radical change in the last decade with a higher emphasis on co-operation and interchange for enterprise decision making and management. This does not just include geospatial data but also geospatial processing (i.e., query and analysis tasks pertaining to geospatial data). This has not only lead to efforts on format and interface standardi-zation efforts but also generated data sharing infrastructures such as Spatial Data Infrastructures (SDIs) and move towards open technology based architectures (OGC). All this can be argued to be interrelated to the trend of increased integration of the information communities.
It is also important to note that information communities enshrine their identity in the way it uses the GI. These may include the GI data components used, the processing that is ap-plied on such data components and the combinations of these. Thus while the cadastral mapping agency usually creates and maintains the land parcel maps, the utilities agencies create and main-tain the utility data in reference to the base map. Power or sewerage utilities can only refer to the base map and not alter it. Beyond this level one can also see how within a power utility company the operations department is allowed to create or alter the utility network data. Departments such as safety or marketing are only able to view this data besides make specialized analysis.
The view of information communities about GI data and processing also varies and this can be easily seen from the granularity at which the data is used. The notion of granularity can be explained on the basis of how people understand geospatial entities as constituted of other geo-spatial entities. For example the marketing department of the power utility company may visual-ize a building or apartment as a customer point the operations department would consider it as a region with access points (say the main switch, meter etc). It is more interesting to note that people belonging to two different information communities at the same time are able to switch between such granularity levels. The dual role of such people imbibes the mapping of what we call semantics of the GI between domains. The important point that can be noted in this example is that human services are employed to ensure information exchange and hence interoperability in some sense. Such human resources who have experience across domains are crucial for or-ganizations and cross department appointments or transfers are intentionally resorted to. This is applicable to non geographic information as well. However one can understand the scale of this problem when we discuss information access across diverse communities that access and share GI in SDIs.
Given such a situation, information communities increasingly look towards technologies to provide access to information. An infrastructure and syntax to provide access to information can be considered the base requirement before one could even start. SDIs and open architectures promoted by XML based data interchange can be considered to such starting steps. We discuss such steps towards what we can claim as interoperability in the next section.
2. What is Interoperability, how do you assess it?
Interoperability is a heavily loaded term and is perhaps best understood in terms of the two constituting words – interoperate and ability. Simply put, it is the ability of something to interoperate with something else. In the context of GI the somethings could be substituted by GI systems, services, applications and other system constituents. Such a wide definition limits speci-ficity in some way but allows the scope of including various issues that result in interoperability failures.
Interoperability is usually the resultant of the cumulative positive abilities of things to work together. Thus MP3 CDs can be played on all MP3 players, and PCs can interoperate with PDA devices. It is not that interoperability is a deterministic property of a system in every case and may also be achieved by introducing additional components such as power adapters em-ployed for electrical devices in different countries.
v So how do we define interoperability at the first place? Wegner defines interoperability in the context of software components as “ability of two or more software components to cooperate despite differences in language, interface, and execution platform” [2]. This definition already accounts for various levels that may contribute to the overall interoperability. The notion of in-teroperation also needs to be understood in the broader context of human beings and systems. This leads to different levels such as purely technical, organizational and societal interoperabili-ties [3].
A further definition can be found in ISO TC204, which defines interoperability as “the ability of systems to provide services to and accept services from other systems and to use the services so exchanged to enable them to operate effectively together.” Both definitions are useful as engineering design principles fail to account for an objective account for what it means to co-operate or operate effectively which leads to notions of partially interoperable systems. So how do you asses the level of cooperation or effectiveness of operation. A formal approach to this problem is still to be achieved and is the focus of current research.
2.1. Syntax vs. Semantics
The issue of interoperability in information science assumes a different dimension since the subjects that we deal with are not entities themselves but symbols of entities in the real world. Thus the country represented in a GIS is not a country but its symbol and has a concept associated to it as shown by the meaning triangle in figure 1. Here, a symbol tank can represent a battle tank or a water tank depending on the concept it is attached to. Therefore a system that finds tank routes cannot interoperate with a system that uses tank in the fish-tank sense.
Figure 1 Meaning triangle based on Ogden & Richards [4]
The situation is very similar to how we as humans interoperate with such diverse cultures and arguably diverse concepts that we carry in our head. The existence of common languages does not necessarily insure interoperability (although absence of common languages would ne-gate the possibility of attaining any kind of interoperability). Open data formats (XML, GML) and standardized protocols (Such as HTTP and SOAP) present the opportunity to communicate through common languages but ensuring that the concepts that underlie our choice of symbols and communication of such concepts remains at a higher level of the interoperability problem. We chose to term the standardization of language and data formats as syntactic interoperability in line with the linguistic sense of components of syntax. The next level of interoperability that ensures the correct use of concepts across systems is the study of semantic interoperability. In our day to day working we use such semantic interoperation based on conceptual integration and the cognitive psychology underlying such human capabilities is an interesting subject and infor-mation science stands to benefit from its theories
2.2. Cost-benefit analysis of interoperability
Let us discuss the significant fact that since human services account for major cost and effort component in the usage of GI products (consider services as the product offered), it would be significant to include automation as a core component of service discovery and matchmaking. It has been reported [5] that for requesting a typical GI product more that 90 percent of the costs are in the area of human services such as training, consultancy, technical support and this is di-rectly related to issues of interoperability.
So what is the value addition proposition and why is it important to have interoperability. Let us assume that there are two services offered in the market place - one offers the road net-work data and the other offers route finding service. It is obvious that by integrating these two services customers can get a information service as ready to use GI product mentioned earlier. There are four scenarios possible in this regard.
Figure 2 Value of Interoperablity
Lastly, one has to look at the investment required for such a proposition. Investments in standardization and implementation of standards are required for syntactic case. Use of Ontolo-gies or other conceptual modeling tools and investments in such knowledge based systems is the investment for the later case. We discuss these in our next section. It is important to note that in-vestments for semantic interoperability
3. Semantic interoperability
We have attempted to explain our notion of the term semantic interoperability which has been frequently used with different and often confusing connotations. A related explanation is given by Doer et al [6] who maintain that information integration at semantic levels consists of mapping between concepts of the two systems or communities. This naturally requires that con-cepts should be formally specified. Gruber [7] defines ontologies as “an explicit specification of a conceptualization” and thus such ontologies can form the basis of the mapping between con-cepts and hence semantic interoperability.
Ontologies have been used in the domain of knowledge representation and artificial intel-ligence from a long time. Conceptualizations that are behind the objects and artifacts that consti-tute the system are domain dependent. These conceptualizations that are expressed as ontologies usually specify objects and their relation to other objects. So how do ontologies help in ensuring semantic interoperability and do they really help? A flat answer to this question is currently there are some efforts to achieve semantic integration based on ontology alignment [8, 9]. Several challenges still exist, ranging from automation of the ontology mapping process, relating concepts on similarity and automated reasoning.
3.1. Web services and semantics
The W3C defines a web service as “software application identified by a URI, whose in-terfaces and binding are capable of being defined, described and discovered by XML artifacts and supports direct interactions with other software applications using XML based messages via Internet-based protocols” [http://www.w3.org/2002/ws/]. The emergence of Service oriented Ar-chitectures (SoA) is an important development by itself but more significantly the limelight on semantic interoperability due to the distributed nature of the SoA has brought major investments into the Semantic Web.
The Semantic Web is defined by the W3C as “an extension of the current Web in which information is given a well-defined meaning, better enabling computers and people to work in cooperation. It is the idea of having data on the Web defined and linked in a way that it can be used for more effective discovery, automation, integration and reuse across various applications. The Web can reach its full potential if it becomes a place where data can be processed by auto-mated tools as well as people” [http://www.w3.org/2001/sw/] Some of the important hurdles of the semantic web are in the context of
Figure 3 The Semantic Wave [Breners & Lee, 2003]
3.2. Semantics of GI
It is important to examine some of the uniqueness of GI in the context of semantics so that these can be taken care while dealing with GI web services and semantic interoperability problems there in.
4. Where do we go from here
After having identified some of the challenges of semantic interoperability specially re-lated to GI we can now look at some of the approaches that are being taken to solve such prob-lems. Again, these are non-exhaustive but only indicative of some directions that are being taken.
Shared vocabularies and ontological mappings: Shared vocabularies form the basis of common ontological commitment and can be found in the knowledge tools of user do-mains/communities. Such domain ontologies can be used to align more specific applica-tion and task ontologies. Domain ontologies, in turn, are expected to be aligned to upper level ontologies. However from the abstract considerations about the distributed nature of knowledge as well as from observation of actual (human) ontology negotiation proc-esses it seems clear that globally agreed-upon conceptualizations are difficult to obtain. Hence there is an increased focus on ontological mappings based on a process of merging and aggregation.
Grounding in cognitive linguistics and measurement theory: The ontological negotia-tion process in humans discussed can be argued to be best understood from cognitive lin-guistics and principles there in such as embodied learning from image schemas [11]. Such schemas provide precognitive basis for formation of human concepts and therefore allow grounding of concepts. Similarly, measurement theory [12] is also important in grounding of GI concepts world which are often based on measurements.
Semantic Reference Systems: The term grounding used above extends the metaphor of referencing things to some point on the ground. Semantic Reference Systems [13] is a similar idea of extending the notion of Spatial Reference Systems that is commonly used in maps and GIS, to the non spatial component of GI that needs semantic referencing.
Figure 4 Reference systems for interpreting geospatial information [14]
All interoperability problems are part of the greater system integration problem and one can only hope that in future technology would provide greater automation towards such integra-tion. This would allow the true extent of GI usage in human society from navigation services to emergency management.
5. References
1. GI usage and Information communities
Users of GI come from different domains and are ever expanding. Traditionally such in-formation resided locally with the users and human interfacing provided the means to usage across domains. Such user domains could be at various levels ranging from industry verticals, cultural-linguistic communities or even departments in the same organization. Such domains can be characterized by the term information communities [1].
Access and usage of GI has undergone radical change in the last decade with a higher emphasis on co-operation and interchange for enterprise decision making and management. This does not just include geospatial data but also geospatial processing (i.e., query and analysis tasks pertaining to geospatial data). This has not only lead to efforts on format and interface standardi-zation efforts but also generated data sharing infrastructures such as Spatial Data Infrastructures (SDIs) and move towards open technology based architectures (OGC). All this can be argued to be interrelated to the trend of increased integration of the information communities.
It is also important to note that information communities enshrine their identity in the way it uses the GI. These may include the GI data components used, the processing that is ap-plied on such data components and the combinations of these. Thus while the cadastral mapping agency usually creates and maintains the land parcel maps, the utilities agencies create and main-tain the utility data in reference to the base map. Power or sewerage utilities can only refer to the base map and not alter it. Beyond this level one can also see how within a power utility company the operations department is allowed to create or alter the utility network data. Departments such as safety or marketing are only able to view this data besides make specialized analysis.
The view of information communities about GI data and processing also varies and this can be easily seen from the granularity at which the data is used. The notion of granularity can be explained on the basis of how people understand geospatial entities as constituted of other geo-spatial entities. For example the marketing department of the power utility company may visual-ize a building or apartment as a customer point the operations department would consider it as a region with access points (say the main switch, meter etc). It is more interesting to note that people belonging to two different information communities at the same time are able to switch between such granularity levels. The dual role of such people imbibes the mapping of what we call semantics of the GI between domains. The important point that can be noted in this example is that human services are employed to ensure information exchange and hence interoperability in some sense. Such human resources who have experience across domains are crucial for or-ganizations and cross department appointments or transfers are intentionally resorted to. This is applicable to non geographic information as well. However one can understand the scale of this problem when we discuss information access across diverse communities that access and share GI in SDIs.
Given such a situation, information communities increasingly look towards technologies to provide access to information. An infrastructure and syntax to provide access to information can be considered the base requirement before one could even start. SDIs and open architectures promoted by XML based data interchange can be considered to such starting steps. We discuss such steps towards what we can claim as interoperability in the next section.
2. What is Interoperability, how do you assess it?
Interoperability is a heavily loaded term and is perhaps best understood in terms of the two constituting words – interoperate and ability. Simply put, it is the ability of something to interoperate with something else. In the context of GI the somethings could be substituted by GI systems, services, applications and other system constituents. Such a wide definition limits speci-ficity in some way but allows the scope of including various issues that result in interoperability failures.
Interoperability is usually the resultant of the cumulative positive abilities of things to work together. Thus MP3 CDs can be played on all MP3 players, and PCs can interoperate with PDA devices. It is not that interoperability is a deterministic property of a system in every case and may also be achieved by introducing additional components such as power adapters em-ployed for electrical devices in different countries.
v So how do we define interoperability at the first place? Wegner defines interoperability in the context of software components as “ability of two or more software components to cooperate despite differences in language, interface, and execution platform” [2]. This definition already accounts for various levels that may contribute to the overall interoperability. The notion of in-teroperation also needs to be understood in the broader context of human beings and systems. This leads to different levels such as purely technical, organizational and societal interoperabili-ties [3].
A further definition can be found in ISO TC204, which defines interoperability as “the ability of systems to provide services to and accept services from other systems and to use the services so exchanged to enable them to operate effectively together.” Both definitions are useful as engineering design principles fail to account for an objective account for what it means to co-operate or operate effectively which leads to notions of partially interoperable systems. So how do you asses the level of cooperation or effectiveness of operation. A formal approach to this problem is still to be achieved and is the focus of current research.
2.1. Syntax vs. Semantics
The issue of interoperability in information science assumes a different dimension since the subjects that we deal with are not entities themselves but symbols of entities in the real world. Thus the country represented in a GIS is not a country but its symbol and has a concept associated to it as shown by the meaning triangle in figure 1. Here, a symbol tank can represent a battle tank or a water tank depending on the concept it is attached to. Therefore a system that finds tank routes cannot interoperate with a system that uses tank in the fish-tank sense.
Figure 1 Meaning triangle based on Ogden & Richards [4]
The situation is very similar to how we as humans interoperate with such diverse cultures and arguably diverse concepts that we carry in our head. The existence of common languages does not necessarily insure interoperability (although absence of common languages would ne-gate the possibility of attaining any kind of interoperability). Open data formats (XML, GML) and standardized protocols (Such as HTTP and SOAP) present the opportunity to communicate through common languages but ensuring that the concepts that underlie our choice of symbols and communication of such concepts remains at a higher level of the interoperability problem. We chose to term the standardization of language and data formats as syntactic interoperability in line with the linguistic sense of components of syntax. The next level of interoperability that ensures the correct use of concepts across systems is the study of semantic interoperability. In our day to day working we use such semantic interoperation based on conceptual integration and the cognitive psychology underlying such human capabilities is an interesting subject and infor-mation science stands to benefit from its theories
2.2. Cost-benefit analysis of interoperability
Let us discuss the significant fact that since human services account for major cost and effort component in the usage of GI products (consider services as the product offered), it would be significant to include automation as a core component of service discovery and matchmaking. It has been reported [5] that for requesting a typical GI product more that 90 percent of the costs are in the area of human services such as training, consultancy, technical support and this is di-rectly related to issues of interoperability.
So what is the value addition proposition and why is it important to have interoperability. Let us assume that there are two services offered in the market place - one offers the road net-work data and the other offers route finding service. It is obvious that by integrating these two services customers can get a information service as ready to use GI product mentioned earlier. There are four scenarios possible in this regard.
- Total Interoperability is assured
- Syntactic interoperability is assured
- Only semantic interoperability is assured
- No Interoperability is assured
Figure 2 Value of Interoperablity
Lastly, one has to look at the investment required for such a proposition. Investments in standardization and implementation of standards are required for syntactic case. Use of Ontolo-gies or other conceptual modeling tools and investments in such knowledge based systems is the investment for the later case. We discuss these in our next section. It is important to note that in-vestments for semantic interoperability
3. Semantic interoperability
We have attempted to explain our notion of the term semantic interoperability which has been frequently used with different and often confusing connotations. A related explanation is given by Doer et al [6] who maintain that information integration at semantic levels consists of mapping between concepts of the two systems or communities. This naturally requires that con-cepts should be formally specified. Gruber [7] defines ontologies as “an explicit specification of a conceptualization” and thus such ontologies can form the basis of the mapping between con-cepts and hence semantic interoperability.
Ontologies have been used in the domain of knowledge representation and artificial intel-ligence from a long time. Conceptualizations that are behind the objects and artifacts that consti-tute the system are domain dependent. These conceptualizations that are expressed as ontologies usually specify objects and their relation to other objects. So how do ontologies help in ensuring semantic interoperability and do they really help? A flat answer to this question is currently there are some efforts to achieve semantic integration based on ontology alignment [8, 9]. Several challenges still exist, ranging from automation of the ontology mapping process, relating concepts on similarity and automated reasoning.
3.1. Web services and semantics
The W3C defines a web service as “software application identified by a URI, whose in-terfaces and binding are capable of being defined, described and discovered by XML artifacts and supports direct interactions with other software applications using XML based messages via Internet-based protocols” [http://www.w3.org/2002/ws/]. The emergence of Service oriented Ar-chitectures (SoA) is an important development by itself but more significantly the limelight on semantic interoperability due to the distributed nature of the SoA has brought major investments into the Semantic Web.
The Semantic Web is defined by the W3C as “an extension of the current Web in which information is given a well-defined meaning, better enabling computers and people to work in cooperation. It is the idea of having data on the Web defined and linked in a way that it can be used for more effective discovery, automation, integration and reuse across various applications. The Web can reach its full potential if it becomes a place where data can be processed by auto-mated tools as well as people” [http://www.w3.org/2001/sw/] Some of the important hurdles of the semantic web are in the context of
- Locating Services (Different terms used for service advertisements and requests)
- Negotiating contracts & communications (Different protocols used for agreements in transactions)
- Service invoking (constructing valid messages based on the published signature of the service)
- Understanding results of the invoked service
- Composing services (constructing composite service based on available services)
Figure 3 The Semantic Wave [Breners & Lee, 2003]
3.2. Semantics of GI
It is important to examine some of the uniqueness of GI in the context of semantics so that these can be taken care while dealing with GI web services and semantic interoperability problems there in.
- GI usually deals with physical locations and entities which exist in the real world. Further, it is also true that such information is often based on measurement values in the real world rather than artificial values.
- The semantics of geospatial information also has a temporal basis and hence a tem-perature, a real-estate price or a bathometry value is relevant to a particular time as it is to a place.
- Social agreements on semantics based on naming of geographic features (physical features such as mountains and lakes as well as man made features such as countries) are also essential part of GI.
- Semantics of these geographic features are also closely linked to the geospatial proc-esses that they participate in. The behavior of geospatial entities with respect to other entities in many ways defines its semantics. For example mountains are relatively dif-ficult to climb (for humans) as compared to hills. However such relations are rela-tively difficult to express in current taxonomy based approaches of ontological speci-fications.
- GI semantics are often associated to vagueness in terms of the granularities at which they are defined. Non-monotonic reasoning capacity in humans usually allows such multiple levels of granularities and selection/amalgamation.
4. Where do we go from here
After having identified some of the challenges of semantic interoperability specially re-lated to GI we can now look at some of the approaches that are being taken to solve such prob-lems. Again, these are non-exhaustive but only indicative of some directions that are being taken.
Shared vocabularies and ontological mappings: Shared vocabularies form the basis of common ontological commitment and can be found in the knowledge tools of user do-mains/communities. Such domain ontologies can be used to align more specific applica-tion and task ontologies. Domain ontologies, in turn, are expected to be aligned to upper level ontologies. However from the abstract considerations about the distributed nature of knowledge as well as from observation of actual (human) ontology negotiation proc-esses it seems clear that globally agreed-upon conceptualizations are difficult to obtain. Hence there is an increased focus on ontological mappings based on a process of merging and aggregation.
Grounding in cognitive linguistics and measurement theory: The ontological negotia-tion process in humans discussed can be argued to be best understood from cognitive lin-guistics and principles there in such as embodied learning from image schemas [11]. Such schemas provide precognitive basis for formation of human concepts and therefore allow grounding of concepts. Similarly, measurement theory [12] is also important in grounding of GI concepts world which are often based on measurements.
Semantic Reference Systems: The term grounding used above extends the metaphor of referencing things to some point on the ground. Semantic Reference Systems [13] is a similar idea of extending the notion of Spatial Reference Systems that is commonly used in maps and GIS, to the non spatial component of GI that needs semantic referencing.
Figure 4 Reference systems for interpreting geospatial information [14]
All interoperability problems are part of the greater system integration problem and one can only hope that in future technology would provide greater automation towards such integra-tion. This would allow the true extent of GI usage in human society from navigation services to emergency management.
5. References
- Bishr, Y., et al.: Probing the Concept of Information Communities - A First Step Toward Semantic Interoperability, in Interoperating Geographic Information Systems (Proceed-ings of Interop'97), M.F. Goodchild, et al., Editors. 1999, Kluwer: pp. 55-71.
- Wegner, P.: Interoperability. ACM Computing Surveys, 28(1):285-287, 1996.
- Harvey, F., et al.: Semantic Interoperability: A Central Issue for Sharing Geographic In-formation. Annals of Regional Science, 1999. 33 (2) pp. 213-232.
- Ogden, C.K. and I.A. Richards,: The Meaning of Meaning. 1946: Harcourt, Brace & World.
- Brox, C. and W. Kuhn.: Demand of Financial Service Providers for GI Products and Electronic Marketplaces. GSDI-7, Bangalore, India. www.gsdi7.org.in, 2004
- Doerr, M. et al.: Towards a Core Ontology for Information Integration. Journal of Digital information, 2003. 4(1).
- Gruber, T.R.: Toward Principles for the Design of Ontologies Used for Knowledge Shar-ing. International Journal of Human and Computer Studies 43(5/6): 907-928, 1993
- Cruz, I. F. et al.: Handling Semantic Heterogeneities Using Declarative Agreements. In ACM GIS 2002, pages 168–174, 2002.
- Kiyoki Y, et al.: A fundamental Framework for realizing semantic Interoperablity in a Multidatabase Environment. Integrated Computer Aided Engineering, 2(1) 3- 20, 1995
- Gärdenfors, P.: How to Make the Semantic Web More Semantic. in Formal Ontology in Information Systems, Proceedings of the Third International Conference (FOIS 2004), A. Varzi and L. Vieu, Editors. 2004, IOS Press, 114: pp. 17-34.
- Johnson, M.: The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason. 1987: The University of Chicago Press.
- Stevens, S.S.: On the Theory of Measurement. Science, 1946.
- Kuhn, W.: Semantic Reference Systems. International Journal of Geographic Information Science (Guest Editorial), 2003. 17(5): pp. 405-409.
- Kuhn, W.: Geospatial Semantics: Why, of What, and How? In Journal of Journal on Data Semantics III, (submitted)