Thursday, 20 January 2011

OpenStreetMap GB - a progress report!

Back in May 2010 ITO World Ltd announced a new online service that compared road names in OpenStreetMap with names released as Open Data by the Ordnance Survey in their OS Locator product in April 2010. We then launched an analysis service which provided statistics for each district/borough in July 2010.

We are now pleased to be able to announce an upgrade to this service which includes a useful thematic map view showing the level of completeness (at the road level) for every district/borough in Great Britain (compared to OS Locator). Here is the overview. Red equals 0-50%, orange 50-75%, yellow 75-95% and blue 95-100%. It is a slippery map and one can zoom into any part of the country.

The district summary report indicates that there are currenrtly 34 districts have greater than 95% road completeness and 171 have over 75% of the roads included (out of a total of 408 districts/boroughs).

As well as thematic view there is still the 'roads missing' view which has been enhanced with administrative boundaries (the thin red lines). The prominence of 'minor errors' as been reduced by always showing them in grey (minor errors are where the only disagreement is in the use of apostrophes - ie Kings Street and King's Street etc).

This information is also available in the main Potlatch editor (use the new Potlatch 2 version to immediate access to the layer).

We have also introduced a page for each district detailing the road missing from OpenStreetMap and also any possible errors in the OS Locator. This page is useful both for mappers wishing to improve the data in their areas and also to authorities wishing to assess the usefulness of OpenStreetMap within their territory. This image shows the page for Cambridge.

This report indicates that in Cambridge there are a total of 66 roads in OS Locator that are missing from OSM and also a total of 7 possible major errors and 4 minor errors in OS Locator itself. We do not currently identify roads that are in OpenStreetMap but which are missing from OS Locator which might also be useful to the OS. Personally I think this looks like 'Big Society' in action!

Here are the possible major errors in OS Locator:
Basset Close should be "Bassett Close"
Chesnut Grove should be "Chestnut Grove"
Kathleen Elliot Way should be "Kathleen Elliott Way"
J J Thompson Avenue should be "J J Thomson Avenue"
Malletts Road should be "Mallets Road"
Northfields Avenue should be "Northfield Avenue"
Winderemere Close should be "Windermere Close"

And the minor errors:
Coldham's Lane should be "Coldhams Lane"
Coldham's Grove should be "Coldhams Grove"
Friar's Close should be "Friars Close"
Parson's Court should be "Parsons Court"

Overall OpenStreetMap now includes 588,000 names roads that match with names in OS Locator (out of a total of 838,000) and of these 140,000 have been added since 1 April 2010 when OS Locator was released. Of course OpenStreetMap is much more than just a road map and has a huge amount of additional detail that is not available from the OS.

For the avoidance of doubt the Ordnance Survey will validate any claims independently for their own reassurance and also to avoid any licensing issues with the OSM CCBYSA and ODBL licenses used by OSM.

For the purposes of this analysis we are using the OS Boundary Line data for districts/boroughs directly to ensure that we have 100% accurate coverage (boundaries are not fully represented in OSM directly yet).

This product has been developed with financial support from the Department of Transport and from the Technology Strategy Board as part of their funding of Ideas in Transit, a five year project to "promote the understanding, awareness and development of user innovations relevant to transport". ITO World Ltd, the Ordnance Survey, the University and the West of England and the University of Loughborough make up the project team.

1 comment:

Simon Hewison said...

Excellent service, but there are a few cases in Wales, where OS has incorrect names in both Welsh and English language, so it needs to pick up not:name:cy and not:name:en as exceptions as well as the normal not:name