Thursday, 3 June 2010

Wrong road names

The Ordnance Survey data that has been released is quite a mixed bag. One of the things I have overlooked is OS Locator. This gives a list of roads with a containing rectangle and some location data, such as settlement and county. The coordinates are in OS northings and eastings of course. Itoworld have taken this data and compared it to OSM data in Britain. They created an overlay showing the locations where OSM and OS road names differ, using the OS rectangle to highlight the area. They have also produced a summary of the number of differences in districts, counties and unitary authorities. As usual with Itoworld's work it is useful and accurate but carries the small penalty of taking a couple of days for changes to be reflected.

I took a look at my local area and immediately found errors. This is not a surprise, my work is not being checked on the ground by anyone else and I am as prone to errors as the next man.

When I looked closely it was clear that I couldn't tell if the OS data was wrong or if there were mistakes in the OSM data without checking it on the ground. I selected a dozen or so fairly close to home to check out.

I found a small residential road I had missed, a couple of tracks that were not on our map, one with a clear name plate I had missed. I found a small industrial site with an apparently public road through it that was missing from OSM. There were a few mis-spellings all on my part when I haven't transcribed from the photograph accurately.

Lastly I found two roads where the OS data had a fault. One was easy, OS call a road in Cottingham Middle Dike Lane, the name board shows (and we know from our local knowledge) it is Middledyke Lane. One photo of the name board and a check that there are no other name boards and we were sure. 

The second was a bit harder. There is a road called Endike Lane that runs from Hull to Cottingham, at least that's how it looks, but the name board in Cottingham says Endyke Lane and the name boards in Hull say Endike Lane. When I mapped the area I assumed that the name changed at the boundary of the city, but now I know that is not quite true. We found the break point by looking at the house numbers, they number from each end towards the join, so 601 Endike Lane is next door to 42 Endyke Lane and the break must lie between them. This is not as OS describes it, but OSM was wrong too, the boundary is about 200m away. I suspect that at sometime in the past the city boundary did lie where the name changes but has moved since.

To prevent the difference being highlighted again, even though we now know that the OSM data is correct, Peter Miller of Itoworld suggests adding a not:name=Middle Dike Lane. I was unhappy at first about adding data to OSM to show this, but now I accept that it is just metadata like a few other widely used tags, and, unlike any others, it is a negative tag which could be useful elsewhere.

All things considered, this new overlay will improve the quality of OSM, especially in areas where lone mappers are working and I welcome it.

No comments: