RDF Mapping Description

The current approach of Linked Geo Data is to store the mapping information in tables together with the OSM database, and use Sparqlify for generating RDF.
The mappings are available at

Mappings.sql

Below information is outdated

Here we describe the mapping rules for converting Open Street Map data into RDF, and the filtering that is performed.

OSM-RDF-Mapping

The current OSM-LGD mappings are configured with this XML file.
Actually this file is a sequence of serialized Java objects. The reason for this is hoped-for flexibility: If the need for a new mapping-type arises, there is a chance that it can be satisfied with a new implementation of a class.

Filtering

The whole Open Street Map dataset is huge. Already in late 2010 the amount of nodes exceeded 1.000.000.000 nodes.
So in order to be performance-wise on the safe side, we have three filters in place:

• The EntityFilter rejects an OSM entity if any of its tags is blacklisted.
• The TagFilter removes individual blacklisted tags from an entity but it does not reject the entity itself.
• And eventually the RelevanceFilter only accepts entities of which any of its remaining tags is explicitly white listed.

Open Street Map entities (nodes and ways) currently need to pass all of these filters in order to end up in the dumps and the SPARQL-Endpoints. Additionally, ways with more than 20 nodes are rejected.

All filters operate on tags and use the same configuration format:
A config file is interpreted as a blacklist, unless it starts with 'whitelist'.
Each further row specifies a pattern of what tags to match.
It has the format

The ± lets one choose between the following two interpretations:

• +: Match any tag with the specified key and values
• -: Match any tag with the specified key, except for those with the given values

The filter files for the latest release are as follows:
EntityFilter
TagFilter
RelevanceFilter