Download presentation
Presentation is loading. Please wait.
Published byMarcus Cross Modified over 7 years ago
1
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 1 LinkedGeoDat a Adding a Spatial Dimension to the Web of Data Sören Auer, Jens Lehmann AKSW Research Group University of Leipzig http://linkedgeodata.org
2
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 2 Structure Introduction & Motivation OpenStreetMap Conversion to RDF/OWL Publishing LinkedGeoData DBpedia Mapping LGD Browser
3
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 3 Introduction & Motivation Real life information integration tasks require spatial knowledge: Offerings of bakeries next door Map of distributed branches of a company Historical sights along a bicycle track LOD cloud contains data sets with spatial features (Geonames, DBpedia, US census, EuroStat), but... they are restricted to popular or large entities like countries, famous places etc. … they lack buildings, roads, mailboxes, trash bins (e.g. Germans like to separate trash),...
4
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 4 OpenStreetMap Goal: create free map of the world Large community project with high growth rate Information quality and density surpasses commercial geo- providers in some regions OSM OpenGL 3D Edit Visualisation: Created by OSM user Ito – see http://www.vimeo.com/2598878http://www.vimeo.com/2598878
5
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 5 OpenStreetMap OpenStreetMap data structure potential crystallization point for spatial Web Data integration OSM Statistics as of June 2009: CategoryOverall AmountAdditions DailyMonthly Growth (last year) users127 thousand20011% Uploaded GPS points 915 million1.6 million10% nodes374 million400 thousand5% ways30 million30 thousand7%
6
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 6 OpenStreetMap (Credits to Iván Sánchez Ortega for this overview.) OpenLayer s (“slippy map”) Potlatch www.osm.org website
7
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 7 OpenStreetMap (Credits to Iván Sánchez Ortega for this overview.) OpenLayer s (“slippy map”) PotlatchJOSM GPX traces, photos¬es WMS services Yahoo! imagery Merkaator www.osm.org website map editing software
8
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 8 OpenStreetMap (Credits to Iván Sánchez Ortega for this overview.) OpenLayer s (“slippy map”) API 0.6 PotlatchJOSM import scripts Geodat a GPX traces, photos¬es WMS services Yahoo! imagery Merkaator www.osm.org website map editing software
9
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 9 OpenStreetMap (Credits to Iván Sánchez Ortega for this overview.) PostGIS OpenLayer s (“slippy map”) Mapnik style- sheets API 0.6 Mapnik + mod_tile cache osm2pgsql PotlatchJOSM import scripts Geodat a GPX traces, photos¬es WMS services Yahoo! imagery map tiles t@hngo server MySQL tiles@home clients Merkaator XAPI Mapnik renderer (tile.osm.org) tiles@home (osmarender) www.osm.org website map editing software
10
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 10 OpenStreetMap (Credits to Iván Sánchez Ortega for this overview.) PostgreSQL backend Planet dump, Planet diffs PostGIS OpenLayer s (“slippy map”) Mapnik style- sheets API 0.6 Mapnik + mod_tile cache osm2pgsql osmosis PotlatchJOSM import scripts Geodat a GPX traces, photos¬es WMS services Yahoo! imagery map tiles t@hngo server MySQL tiles@home clients Merkaator XAPI Mapnik renderer (tile.osm.org) tiles@home (osmarender) www.osm.org website map editing software
11
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 11 OpenStreetMap Data Model Nodes: points on earth with lat/long values Ways: ordered sequences of nodes Relations: groupings of nodes and ways name = Zoo Leipzig tourism = zoo wikipedia = http://en.wikipedia[...]Leipzig_Zoo created_by = Potlatch 0.10f + key / value pairs }
12
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 12 Conversion to RDF/OWL Node, way, relation as subclasses of wgs84:SpatialThing Use of wgs84:long, wgs84:lat Three categories of attributes: ● Classification attributes, e.g. highway = motorway ● Description attributes, e.g. created_by = JOSM ● Data attributes, e.g. postcode = 01187 Converted to class hierarchy, object properties, data properties, respectively: 500 classes, 50 object properties, 15000 data properties
13
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 13 Publishing LinkedGeoData We used Triplify (http://triplify.org) - light- weight, simplistic, allows to integrate relational and RDF datahttp://triplify.org Data is stored in relational database for Linked Data publication RDF dumps available at http://linkedgeodata.org/Datasetshttp://linkedgeodata.org/Datasets SPARQL endpoint http://lod.openlinksw.com/sparqlhttp://lod.openlinksw.com/sparql Statistics: 3 billion triples (remaining LOD cloud contains 4.7 billion triples), 350 million nodes, 30 million ways – 122 Mio triples ”interesting” about 12 million POIs
14
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 14 Publishing LinkedGeoData Natural entry point for geo data is neighborhood of a particular point: Other REST interfaces available: Points in a circular area Points in a circular area belong to a certain class Points in a circular area having a certain property value Points in rectangle area are retrieved (efficiently indexed) and restricted to circle using Haversine formula http://LinkedGeoData.org/triplify/near/48.213,16.359/100 0 http://LinkedGeoData.org/.../48.213,16.359/1000/class/pub http://LinkedGeoData.org/.../48.213,16.359/1000/postal_code=108 0 latitude longitude radius
15
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 15 DBpedia Mapping DBpedia is a central interlinking hub in the Web of Data By connecting LGD and DBpedia via owl:sameAs, we also connect it to a variety of other data sets Existing links in OSM to Wikipedia pages for bootstrapping the mapping process DL-Learner machine learning tool identified schema- level mappings, e.g. for cities: (dbpedia-owl:City or umbel-sc:City) is close to (lgd:City or lgd:Town or lgd:Village or lgd:Suburb)
16
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 16 DBpedia Mapping selected entities in DBpedia having latitude and longitude For each such point try to find a match using three matching criteria: Type information (schema matching) Spatial distance Name similarity All those necessary to enable an efficient and precise mapping
17
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 17 DBpedia Mapping – Step By Step Given a DBpedia point, query LGD points within type specific maximum distance Compute spatial score Compute name similarity: name, name_en, name_int rdfs:label Jaro string metric Combine both scores owl:sameAs match is highest point above threshold
18
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 18 DBpedia Mapping Type#Matche s Rat e city4572 9 70.9 % railway station 92 9 24.8 % universit y 21 0 13.3 % schoo l 148 3 38.4 % airpor t 64 9 8.4 % lak e 101 4 22.1 % countr y 16 0 20.1 % islan d 31 3 29.8 % mountai n 147 5 24.5 % river67 7 32.0 % lighthous e 2525 4.3 % stadiu m 34 6 17.0 % >50 000 matches, mostly cities Covers 53.8% of DBpedia geo entities of considered types Reasons for not finding matches: Resource does not exist in LGD Resource not sufficiently described (type, name) Errors in DBpedia
19
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 19 LGD Browser Editable description FacetsZoom levels OSM loginPOI summary Demo
20
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 20 Efficient Spatial Indexing Naive LGD querying would be extremely slow LGD uses Quadtile indexing Aggregate counts for classes and property-value combinations are precomputed in discrete hypercubes for 18 zoom levels
21
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 21 Future Work Integrate with more knowledge bases, e.g. Geonames Connecting to other domains, e.g. LGD and marine sciences (FishBase, SeaLifeBase, AquaMaps) Refine LGD schema, potentially with community help and feed back observations to OSM community Continue work on LGD browser (synchronisation, user interface) Apply best practices to be discussed at NeoGeoVoCamp Washington DC
22
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 22 Take-Home Messages OpenStreetMap immensely successful project for collaboratively creating free spatial data Community uses key value structures, which provide a rich source of information Key strength: broad coverage Established mapping to DBpedia Facet-based LGD Browser provides an interface for OSM/LGD, which highlights its structural aspects
23
Universität Leipzig ▪ Agile Knowledge Engineering and Semantic Web (AKSW) Authors: Sören Auer, Jens Lehmann Slide 23 The End Thanks for your Attention! More details and demos at the poster session this evening.
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.