Thursday, 13 September 2012

OS and OSM road names

OS Locator is useful. As I have said before it helps OSM mappers in Great Britain find places where the names of roads we have mapped disagree with the names Ordnance Survey have. The reasons the names differ are many and vary across different types of road. Many country roads just don't have a name board, so if we use the OS name there can be no difference  - except that the names on OS StreetView are sometimes different from the names on OS Locator. This might imply more errors being possible where all the roads are named. There are typographical differences such as names with or without a space (Tweendykes Lane vs Tween Dykes Lane) or use of apostrophes. Sometimes the name is just completely wrong. Just occasionally a road can have two or more name boards that do not agree, but I've only seen a small difference such as use or not of an apostrophe.

OS do not provide the definitive name, that is the local council that the road falls into. Where a road straddles two authorities this can cause a problem: a road called Endyke Lane in East Yorkshire turns into Endike Lane as it crosses into Hull, though now not exactly on the boundary.

I believe the name board on the ground to usually be correct, after all if the board was wrong people on that street are likely to ask questions of their council and it may well get changed.

I saw a comment earlier today on twitter from Tom Chance (@tom_chance) about the discrepancies in Southwark and they looked to be at a similar level to East Yorkshire and Hull that I have worked on. I thought I'd take a look, and ended up examining some some from the very useful ITOWorld OSM Analysis page. I took the twenty nine areas that were 100% complete and worked out the percentage of name anomalies for each area. The graph of the results is below:

The average percentage of anomalies is 2.65%. The idea that there are no anomalies in Taunton Deane seems odd to me, though possible.

This looks like a spread of differences that might reflect the fact that different surveyor work in different areas and everyone makes mistakes. Has an OS surveyor misread or misrecorded or mistranscribed a name? I feel that many of the anomalies will have been rechecked on the ground by OSM surveyors, in case they were wrong, so in general I trust OSM more than OS. I hope that no one just uses the OS names in the belief they are always right.  There is also a very small chance that both OS and OSM have misrecorded a name in the same way so they agree (no anomaly) but are both wrong. 


Gregory Williams said...

I think that there's probably some variation on how people in OSM handle apostrophes and whether OS have a different version of the name. I use not:name to point out where there's a difference of just an apostrophe, but I think some others just put the alternative in an alt_name tag.

It'd be interesting to see the same analysis where differences of just an apostrophe are ignored.

Anonymous said...

For some areas people just copy from OS and import them directly.

You can see this clearly for a few areas. Where they are at 100% and if you look at and see the hundreds of entries a day they are not from a real life survey.

I imagine Portsmouth at the moment is being mapped not from ground surveys.

I think when an area is being mapped properly 5% or more roads could be different.

Anonymous said...

Doh I meant Plymouth

Chris Hill said...

If people want a copy of OS why not just use OS OpenData and not add their errors to OSM.

Anonymous said...

I understand that but I am pointing out that people do do it. I didn't do those but I am not against it either. Even in my local area with many active mappers there was an area with hundreds of roads that was not touched years on. It was eventually added by someone local via OS info. Nothing was check on the ground but I made the map more complete.

If no-one is physically going to map by hand/on the ground then at least this gets roads on the map.

There were many debates about importing directly from OS when the data became available. And still might be done for some of the areas that have no active mappers and haven't been touched in years.

Even if all the roads were added from OS Opendata. It will not be the same as just OS OpenData, it will be OS+OSM with all the road mapped to a better/reasonable degree.

Currently there are still a lot of roads not named. And most of the roads added are likely from armchair mappers with OS openview data and not on the ground. I think it is naive to think otherwise. Without people liek that many areas would not be over 50% complete instead of 100% complete.

Chris Hill said...

The quality of the OSM map data is an interesting point. Some would say that having the names from OS improves the quality, some would say that incorporating the errors from OS makes the quality worse and that empty space is better than poor quality. There are strongly held views on both sides.

I want OSM to be the best it can be. Adding unchecked imports means errors will be imported too, but OSM mappers make mistakes or cut corners too. The important point is that it is editable, so when anyone sees something wrong they can fix it, wherever it came from.