The document describes a Geographic Knowledge Base (GKB) that was created to integrate geographic and network knowledge from different sources and represent it as ontologies. The GKB combines hierarchies of geographic features from various sources using an algorithm to identify common instances and merge the hierarchies. Rules are then used to assign geographic scopes to websites based on matching their domain names to feature names in the GKB. Applications that have used the GKB include named entity recognition of geographical references and assigning documents to geographical scopes for information retrieval.

1. A Geographic Knowledge Base for Semantic Web Applications Marcirio Silveira Chaves Mário J. Silva Bruno Martins 20º Brazilian Symposium on Databases - SBBD 2005 Uberlândia - MG Linguateca www.linguateca.pt

2. Motivation/Context GKB - Geographic Knowledge Base Geographic Network Information exported as ontologies Geographic-aware Semantic Web applications GREASE – Geographic Reasoning for Search Engines

3. Presentation Structure Conceptual Design of GKB Knowledge Integration Using Geographic Knowledge in GKB GKB as an Ontology Statistics of the Ontologies Created Applications using GKB Final Remarks

4. Information Sources used by GKB Geo-Administrative and Geo-Physical Domain Administrative Postal Gazetteers Wikipedia Network Domain FCCN Web domains Web sites

5. Architecture of GKB

6. Feature concept in GKB A meaningful object in the selected domain of discourse [ISO19109]. Ex.: countries, cities and localities

7. Conceptual Design of GKB GKB meta-model

8. Presentation Structure Conceptual Design of GKB Knowledge Integration Using Geographic Knowledge in GKB GKB as an Ontology Statistics of the Ontologies Created Applications using GKB Final Remarks

9. Knowledge Integration in GKB GKB hierarchy from different information sources Algorithm: It searches the lowest common features types in both hierarchies If it holds, it identifies the common instances between the hierarchies Once the common instances are identified, it goes up the hierarchy and searches for the lowest common ancestor It verifies the distance (in number of relationships partOf) between the common instances of the features types and its ancestors. The ancestor, which has the small distance up to the common instances is merged through a relationship partOf with the ancestor in the another hierarchy. The existing relationships in both hierarchies are maintained.

10. Knowledge Integration in GKB GKB hierarchy from different information sources H1 Norte Grande Porto Tâmega Matosinhos Vila Nova de Gaia Penafiel NUT2 NUT3 MUNICIPALITY MUNICIPALITY H2 Porto Matosinhos Vila Nova de Gaia Penafiel DISTRITO

11. Knowledge Integration in GKB GKB hierarchy from different information sources H1 Norte Grande Porto Tâmega Matosinhos Vila Nova de Gaia Penafiel NUT2 NUT3 MUNICIPALITY MUNICIPALITY H2 Porto Matosinhos Vila Nova de Gaia Penafiel DISTRITO

12. Knowledge Integration in GKB GKB hierarchy from different information sources H1 Norte Grande Porto Tâmega Matosinhos Vila Nova de Gaia Penafiel NUT2 NUT3 MUNICIPALITY MUNICIPALITY H2 Porto Matosinhos Vila Nova de Gaia Penafiel DISTRITO

13. Knowledge Integration in GKB Merged Hierarchy Norte Grande Porto Porto Tâmega Penafiel Matosinhos Vila Nova de Gaia

14. Presentation Structure Conceptual Design of GKB Knowledge Integration Using Geographic Knowledge in GKB GKB as an Ontology Statistics of the Ontologies Created Applications using GKB Final Remarks

15. Using Geographic Knowledge in GKB Geographic scopes www.cm-lisboa.pt Lisboa (municipality) Rules New relationships and knowledge Description Logics (DLs) Geo domain Names composed of multiple words are represented in different ways Network domain Names of URLs are decomposed by the correspondent domain division

16. ABox in DLs for the: municipality of Santiago do Cacém geoFeatureName (270,“santiago do cacem”) geoFeatureName (270,“santiag oc acem”). geoFeatureName (270,“santiago -do- cacem”). geoFeatureName (270,“santiag o-c acem”). geoFeatureType (270,“CON”). web site: www.cm- santiago-do-cacem.pt netSiteSubDomain (33684,“www”). netSitePrefix (33684,“cm”). netSiteDomainToken (33684,“santiago-do-cacem”). netSiteTLD (33684,“pt”). Using Geographic Knowledge in GKB

17. Terminology Description (TBox in DLs) Municipalities hasScope(idN,idG)   netSiteDomainToken (idN,X)  ((  netSitePrefix (idN,“cm”)   netSitePrefix (idN,“mun”))   geoFeatureType (idG,“CON”)   geoFeatureName (idG,X). Using Geographic Knowledge in GKB

18. Ex.: hasScope(idN,idG)   netSiteDomainToken (idN,X)  (  netSitePrefix (idN,“cm”)   netSitePrefix (idN,“mun”))   geoFeatureType (idG,“CON”)   geoFeatureName (idG,X). netSiteDomainToken (33684, “santiago-do-cacem”). netSitePrefix (33684, “cm”). geoFeatureType (270, “CON”). geoFeatureName (270, “santiago -do- cacem”). New knowledge: hasScope( 33684 , 270 ). Using Geographic Knowledge in GKB

19. Rule-based assigned scopes by GKB to sites of Portugal Using Geographic Knowledge in GKB Scopes extended to the web pages under each one of the sites of matching subdomains 105 (26%) 402 high schools 55 (36%) 152 training centers 124 (6%) 1955 basic schools 124 (41%) 300 freguesias 261 (90%) 288 municipalities 17 (52%) 33 distritos # of matches # of sites Site Type

20. Presentation Structure Conceptual Design of GKB Knowledge Integration Using Geographic Knowledge in GKB GKB as an Ontology Statistics of the Ontologies Created Applications using GKB Final Remarks

21. GKB as an Ontology <gn:Geo_Feature rdf:ID=&quot; GEO_238 &quot;> <gn:geo_id> 238 </gn:geo_id> <gn:geo_name xml:lang=&quot;pt&quot;> Porto </gn:geo_name> <gn:geo_type_id rdf:resource=&quot; #CON &quot;/> <gn:info_source_id rdf:resource=&quot; #INE &quot;/> <gn:related_to> <rdf:Bag> <rdf:li> <gn:Geo_Relationship> <gn:rel_type_id rdf:resource=&quot; #PRT &quot;/> <gn:geo_id><rdf:Bag> <rdf:li rdf:resource=&quot; #GEO_130 &quot;/> <rdf:li rdf:resource=&quot; #GEO_3967 &quot;/> </rdf:Bag></gn:geo_id> </gn:Geo_Relationship> </rdf:li> <rdf:li> <gn:Geo_Relationship> <gn:rel_type_id rdf:resource=&quot; #ADJ &quot;/> <gn:geo_id> <rdf:Bag> <rdf:li rdf:resource=&quot; #GEO_127 &quot;/> <rdf:li rdf:resource=&quot; #GEO_156 &quot;/> <rdf:li rdf:resource=&quot; #GEO_162 &quot;/> <rdf:li rdf:resource=&quot; #GEO_331 &quot;/> </rdf:Bag> </gn:geo_id> </gn:Geo_Relationship> </rdf:li> </rdf:Bag> </gn:related_to> <gn:population> 263131 </gn:population> </gn:Geo_Feature> Geo-Net-PT01

22. Statistics of the Ontologies Created 12,291 (99,99%) 417,739 (99.92%) # of features without adjacent 11,819 (96,14%) 417,867 (99.95%) # of features without equivalent 12,045 (97,98%) 374,349 (89.54%) # of features without descendants 1(0.00%) 3 (0.00%) # of features without ancestors 6.5 3.54 Avg. adjacent features per feature with adjacent 3.82 1.99 Avg. equivalent features per feature with equivalent 475.44 10.56 Avg. narrower features per feature 1.07 1.0016 Avg. broader features per feature 13 (0.10%) 1,132 (0.27%) # of adjacency relationships 2,501(20,40%) 395 (0.09%) # of equivalence relationships 12,245 (99,89%) 418,340 (99.83%) # of part-of relationships 12,258 419,867 # of relationships 12,293 418,065 # of features World Portugal Statistic

23. Presentation Structure Conceptual Design of GKB Knowledge Integration Using Geographic Knowledge in GKB GKB as an Ontology Statistics of the Ontologies Created Applications using GKB Final Remarks

24. Applications using GKB NERC tool for recognizing geographical references in text Classification tool for assigning documents to a corresponding geographical scope Information retrieval interface for geographical queries

25. Applications using GKB

26. Final Remarks A domain-independent model for storing geographic and network knowledge Sharing of the collected knowledge as formal ontologies Geo-Net-PT01 : The first public geographic ontology of Portugal - http://xldb.fc.ul.pt/geonetpt Future work Augmenting the knowledge in GKB with geographic entities extracted from the texts of the Portuguese Web