Wikipedia defines building as "a man-made structure with a roof and walls standing more or less permanently in one place" and "intended for human use or occupation" [http://en.wikipedia.org/wiki/Building]. Wikipedia also suggests examining http://en.wikipedia.org/wiki/Nonbuilding_structure for examples of non-building structures, where are listed the Eiffel Tower, the Golden Gate Bridge, roller coasters, water towers, gates, fortifications, and transmission towers.
However, DBpedia includes RollerCoaster, WaterRide, Gate, MilitaryStructure, Treadmill, and WaterTower as subclasses of Building. Each of these are either explicitly listed as non-building structures or are obviously non-building structures given the exclusions in Wikipedia. It turns out that many of the classes that are subclasses of Building in the DBpedia ontology are either disjoint from Building or are not truely subclasses of Building. Similar problems affect ArchitecturalStructure. The definition of architectural structure is unclear, but neither public transit systems nor parks are architectural structures at all, except possibly in very unusual cases, even though these are subclasses of ArchitecturalStructure in the DBpedia ontology. Here is my in-depth analysis of ArchitecturalStructure and its subclasses, building on the observations above. I have used the appropriate pages from English Wikipedia and instances from DBpedia 3.9 to back up my arguments. ([NC] ([NEC]) means no (English) comment on the class.) ArchitecturalStructure [NC] Building is a subclass of Architectural Structure. However, the Infobox Wikipedia defines building as "a man-made structure with a roof and walls standing more or less permanently in one place" and "intended for human use or occupation" [http://en.wikipedia.org/wiki/Building]. Wikipedia also suggests examining http://en.wikipedia.org/wiki/Nonbuilding_structure for examples of non-building structures, where are listed the Eiffel Tower, the Golden Gate Bridge, roller coasters, water towers, gates, fortifications, and transmission towers. However, DBpedia includes RollerCoaster, WaterRide, Gate, MilitaryStructure, Treadmill, and WaterTower as subclasses of Building. Each of these are either explicitly listed as non-building structures or are obviously non-building structures given the exclusions in Wikipedia. It turns out that many of the classes that are subclasses of Building in the DBpedia ontology are either disjoint from Building or are not truely subclasses of Building. Similar problems affect ArchitecturalStructure. The definition of architectural structure is unclear, but neither public transit systems nor parks are architectural structures at all, except possibly in very unusual cases, even though these are subclasses of ArchitecturalStructure in the DBpedia ontology. Here is my in-depth analysis of ArchitecturalStructure and its subclasses, building on the observations above. I have used the appropriate pages from English Wikipedia and instances from DBpedia 3.9 to back up my arguments. ([NC] ([NEC]) means no (English) comment on the class.) ArchitecturalStructure [NC] Building is a subclass of Architectural Structure. However, the Infobox building is too broadly used in English Wikipedia, e.g., http://en.wikipedia.org/wiki/Eiffel_Tower. Proposal: Modify the mapping for Infobox building to check for special cases, such as building_type including "tower", and map them into the appropriate classes. This is the most complex proposal here because of the overuse of the building infobox. [NC] AmusementParkAttraction is excluded implicitly from building. There are no instances of this class that I could find. Proposal: Move the class to be a direct subclass of ArchitecturalStructure. RollerCoaster is excluded explicitly from building. No further change is needed for this class. [NC] WaterRide is excluded implicitly from building. No further change is needed for this class. [NC] Arena is excluded explicitly from building, presumably because of open-air arenas such as http://nl.dbpedia.org/resource/Stadio_Giuseppe_Meazza. Arena is populated directly from the nl:Infobox stadion. Proposal: Move the class to be a direct subclass of ArchitecturalStructure and make no changes to the mapping rule. [NC] Gate is excluded implicitly from building. Proposal: Move the class to be a direct subclass of ArchitecturalStructure. HistoricBuilding needs no change. [NC] MilitaryStructure is not a subclass of building, as it includes fortifications. The class is also used for military bases, which are not even architectural structures, e.g., http://dbpedia.org/resource/Camp_David and http://dbpedia.org/resource/Guantanamo_Bay_Naval_Base. The pages incorrectly use the Infobox military structure. Proposals: Move the class to be a direct subclass of ArchitecturalStructure. Fix mappings for this infobox to check for type Military base, and then instead use type Place. The comment on the class needs to be changed to conform with architectural structure. Mill is not a subclass of building as many mills have no enclosure and are thus are not buidings. The only mills listed are in the Dutch Wikipedia, and are mostly buildings or at least building-like. One exception is http://nl.dbpedia.org/resource/De_Meent_(Langerak). Proposal: Make this class a direct subclass of ArchitecturalStructure. Even though most mills currently in DBpedia could be considered to be buildings, there will be very little lost here. [The comment needs to be changed to broaden to its actual use.] Treadmill, Watermill, WindMotor, and Windmill should remain as subclasses of Mill and thus need no changes besides being removed from Building. ReligiousBuilding the physical presence of many religious congregations are only parts of buildings, but it appears that in DBpedia only includes those meeting places for religious congregations that are buildings. Monastery is not a subclass of building because many monasteries include multiple buildings and other structures, and are thus not even architectural structures, e.g., http://fr.dbpedia.org/resource/Gndevank and http://nl.dbpdia.org/resource/Makarjevklooster. Instances of this class are populated from Infobox Édifice religieux, checking for type Monastère. Dutch instances of this class are populated directly from nl:Infobox klooster. Proposal: Make this a direct subclass of Place. Some information will be lost for monasteries that are single buildings, but this is a minor loss. Abbey is not a subclass of building because some abbeys include multiple buildings and other structures, and are thus not even architectural structures, e.g., http://fr.dbpedia.org/resource/Abbaye_de_Lucedio Proposal: Make this a direct subclass of Place. Some information will be lost for abbeys that are single buildings, but this is a minor loss. Church is not a subclass of Building. "A Church is a religious institution, place of worship, or group of worshipers, usually Christian" http://en.wikipedia.org/wiki/Church. However, all (or almost all) churches in DBpedia are religious buildings, mostly free-standing, as only important Churches have been placed in DBpedia. Proposal: Do not make any change, except putting a comment in the DBpedia ontology that this class is for church buildings only. It would be better to rename this class as something like "Church_(building)", but I do not know if this change is worth it. [NC] Mosque is in the same situation as Church. Synagogue is in the same situation as Church. Temple is only populated in Japanese DBpedia, and is likely in the same situation as Church. [NC] Shrine is not a subclass of Building as almost all shrines are not buildings, being simple and small. Proposal: Make this a direct subclass of ArchitecturalStructure. Skyscraper needs no change. [NC] Tower is not a subclass of Building, as many towers have no interior space. Proposal: Move the class to be a direct subclass of ArchitecturalStructure. [The comment needs to be fixed.] Lighthouse is not a subclass of Building, even though many lighthouses are buildings or building-like. "A lighthouse is a tower, building, or other type of structure" [http://en.wikipedia.org/wiki/Lighthouse]. Proposal: Leave as as direct subclass of Tower, which will move it out of Building. [NC] WaterTower is excluded explicitly from building. Proposal: Leave as as direct subclass of Tower, which will move it out of Building. Venue is not a subclass of Building as many venues are open air, e.g., http://dbpedia.org/resource/Hollywood_Bowl, or parts of buildings. Venue is populated directly from Infobox venue. Proposal: Move the class to be a direct subclass of ArchitecturalStructure. [NC] Theatre is not a subclass of Building as some theatres are open air, e.g., http://dbpedia.org/resource/Delacorte_Theater, and other theatres are parts of larger buildings or structures e.g., http://dbpedia.org/resource/Circle_in_the_Square_Theatre. Proposal: Leave as a direct subclass of Venue, which will move it out of Building. [NC] Casino, Castle, Factory, Hospital, Hotel [NC], Museum [NC], Prison [NC], Restaurant [NC], ShoppingMall [NC] are all functional facilities, which are often either complexes of buildings and other structures, e.g. http://dbpedia.org/resource/Toronto_General_Hospital, or parts of buildings or other structures, e.g., http://dbpedia.org/resource/Casino_Lisboa,_Macau__Robuchon_á_Galera__1 Proposal: Move all these to be direct subclasses of Place. Dam needs no change. [The comment needs to be adjusted.] Garden is not a subclass of ArchitecturalStructure as many gardens are just plantings of flowers with no associated structure. Proposal: Move this class to be a direct subclass of Place. [NC] Infrastructure needs no change [NC] Airport is questionable as a subclass of ArchitecturalStructure, but probably needs no change. [NC] LaunchPad [NC], Lock [NC], PowerStation [NC], and NuclearPowerStation [NC] need no change. RouteOfTransportation [NC] is questionable as a subclass of ArchitecturalStructure, but probably needs no change. Bridge needs no change. [NC] PublicTransitSystem is not a subclass of RouteOfTransportation or even ArchitecturalStructure, e.g., http://dbpedia.org/resource/Massachusetts_Bay_Transportation_Authority. Proposal: Make this class a direct subclass of Organisation. [NEC] RailwayLine, Road [NC], and RoadJunction [NC] need no change. [The commment on RailwayLine probably needs to be adjusted.] Tunnel is not a subclass of RouteOfTransportation as not all tunnels are routes of transportation, e.g., http://dbpedia.org/resource/Tunnel_of_Eupalinos. Proposal: Make this class a direct subclass of ArchitecturalStructure. RailwayTunnel [NC], RoadTunnel [NC], and WaterwayTunnel [NC] are all subclasses of RouteOfTransportation. Proposal: Make these classes subclasses of both Tunnel and RouteOfTransportation. Station, MetroStation [NEC], and RailwayStation [NC] need no change. NoteworthyPartOfBuilding needs no change. Park is not a subclass of ArchitecturalStructure as many parks are mostly natural areas, e.g., http://dbpedia.org/resource/Deception_Pass. Proposal: Make this class a direct subclass of Place. [NC] ------------------------------------------------------------------------------ Learn Graph Databases - Download FREE O'Reilly Book "Graph Databases" is the definitive new guide to graph databases and their applications. Written by three acclaimed leaders in the field, this first edition is now available. Download your free book today! http://p.sf.net/sfu/NeoTech _______________________________________________ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
