Missing Roads

Posted by ca_hoot on 15 July 2012 in English (English)

Just a quick reminder for UK (or anyone else of course) mappers of the excellent ITO - OSM Analysis Map, activity seems to have tailed off a little bit recently with what seems like only me and a few others spending time adding missing UK roads in.

The recent licence change has removed around 3000 non ODbL compliant roads (with some areas to be run again, so this is sure to increase) according to the latest ITO update. Lets make a big push to get them all added back in!

Currents stats are:

  • Roads Missing from OSM (major) 91,799
  • Roads Missing from OSM (minor) 3,971
  • UK Percentage Complete 89.20%
  • Missing change last day 3,038
  • Missing change last week 2,168
  • Missing change last month -5,059

For anyone that doesn't know the ITO Analysis Map compares UK roads to the OS Locator database from Ordnance Survey, which then can be shown as a layer in JOSM or Potlatch so missing roads can easily be added in. Wiki Page

[Update] See map link for current redaction bot mess in Salford/Manchester.

[Update (again)] To address controversy in the comments. Obviously this kind of data should be used with caution! Don't destroy anyone else's work or rename streets that have been previously surveyed. My main reason for making this post was to highlight how ITO could be used post-redaction bot to identify missing streets etc. Ground surveys are always preferable. Sorry for any confusion.

Location: 53.554, -2.285

Comment from Vclaw on 15 July 2012 at 11:08

Have you surveyed any of these roads, to see if the names are actually correct?

Hide this comment

Comment from paulbiv on 15 July 2012 at 11:35

Can't speak for all, but for me the ITO map (and others, there's musicalchairs as well) is a prompt to go out and check. Where the Ordnancy Survey has got it wrong we tag the way not:name with the OS version after putting in the correct one.

The OS makes lots of mistakes as the same character as OSMers who audio map do - putting in likely spellings rather than correct ones. Sometimes they'll put in a planned name from the council that turns up slightly different on the street signs - we map what's on the ground.

Hide this comment

Comment from ca_hoot on 15 July 2012 at 11:43

@ Vclaw Yes some are surveyed, some are not. The fact of the matter is OS Data is generally accurate in areas I have looked at that have been surveyed on foot as well, at an estimate less than 1% errors.

We would need hundreds of thousands more mappers to survey every road in the country. Currently there is around 3000 active on any given day in the entire world, that expectation seems unrealistic to me, after all we are all doing this as a hobby rather than a full time job.

I'd rather have a tiny amount of naming errors than no road at all. Street signs can also be wrong. Data can always be updated to correct any errors.

All named roads that I create (that are not surveyed) are tagged with source and source:name. So there is no doubt about where the information is from and the relative accuracy of it.

No doubt ideally every street and every building and every sign post and every path and every tree should be surveyed, but honestly I'd rather have a usable map within this century.

Hide this comment

Comment from Vclaw on 15 July 2012 at 13:17

You are changing road names previously added by others, which were probably from a survey and/or local knowledge. What you makes you sure those other mappers are wrong?

And by doing this it makes it harder to find actual surveying errors, as the area will now show as 'correct' on tools comparing it to the OS data.

If you want a 'complete' road map you can use the OS OpenData already. There is no need to to dump it all into OSM.

Hide this comment

Comment from ca_hoot on 15 July 2012 at 14:14

@ Vclaw I think you are misunderstand, I don't change any names that are tagged with source:survey or local_knowledge. 99% of roads I tag are source:OS_StreetView without a name tag or do not exist at all. Street view does not include the names of small side streets.

I'm very careful not to destroy other people's work. When I said some are surveyed, I mean't some have been surveyed by me. Not that I change the names of surveyed roads because Locator might say something different.

As for the data being available already, that is entirely true, which is exactly why we shouldn't be wasting time surveying roads that have already been surveyed by someone else largely accurately.

I would go further and suggest that all of this should have been dumped into OSM long ago just like the tiger data, and then checked. Rather than the opposite way around.

If OSM is to become even more popular surely it needs to offer something that other maps don't have, which is detail and local knowledge. But a map with highly detailed pockets around where mappers live but in other areas doesn't even have roads is a useless map. Which is why I'm all for using this data, then mappers can concentrate on adding the things that make this map so much better than any other.

Hide this comment

Comment from ca_hoot on 15 July 2012 at 14:47

@ Vclaw. I have updated my post to reflect your concerns.

Hide this comment

Comment from chriscf on 18 July 2012 at 01:46

What if they're not tagged with "survey" or "local_knowledge" as a source? Are you going to replace possibly correct data with possibly incorrect data?

Please don't use unhelpful measures such as some made-up "completeness" figure. It's a statistic useful for comparison, nothing more. We have some names wrong, the OS also have some names wrong. I had one silly person around me trace roads from StreetView and add the names from Locator. They denied it, however, there were a fair number of things they added which couldn't have come from anywhere else (typos preserved, names with no ground presence and no residential property, non-existent roads added).

Hide this comment

Comment from ca_hoot on 18 July 2012 at 11:26

@chriscf While ground surveys are preferable, OSM is not 'the garmin club'. Potential data sources are wide and varied, although ultimately, of course your eyes are the best source, they are also far from accurate, and consumer GPS devices are extremely inaccurate.

There is no rule that says 'everything you add to this map has be seen with your own eyes first on the ground'.

Tracing is perfectly acceptable and to be encouraged (in a considerate way of course). Why anyone might deny that they did it I have no idea. This is a community, we are collaborating, tracing some roads from StreetView is helping the project not, as you seem to think, hindering it.

These kind of highly negative attitudes will drive away potential mappers, not encourage them. I have made it perfectly clear any mapping should be done with thought and consideration for others, ground based or otherwise. (on a side note I sincerely hope your edits such as 'remove blind mapping' and your seeming huge number of reverts had valid reasons. Because using StreetView or Bing as a source is not one, unless you know something different on the ground)

Hide this comment

Comment from Rovastar on 18 July 2012 at 12:12

Many old school mappers have already forced out newer mappers (and existing) that use sources. There holy than thou attitude is problematic to the community.

Personally I try and survey on the ground and use it with OS locator in areas heavily mapped and/or I am local too.

But armchair mapping for other areas that are not mapped by local and not likely to be mapped. The around 50% coverage on OSM haven't had anyone map on the ground for...well ever.

It is unrealistic to wait another decade or so until we get a mapper in the area.

Hide this comment

Comment from ca_hoot on 18 July 2012 at 12:38

@rovastar I completely agree, one of my main reasons for moving on from mapping my local area via survey and Bing etc, is that I have already mapped it to a very high degree of detail. So I moved onto the north of England and areas that had little or no mapping and hadn't been touched for years... adding missing features using Bing and the various OS sources. How this can be a bad thing I have no idea.

The idea that because a feature is actually in the database then nobody would bother to go out and survey it is just not true. Once mappers in that particular area join up it will be easy to see where that data has come from and correct anything they know is wrong.

I have a feeling this all stems from the fact that a certain subset of mappers that have been here a long time and spent a long time surveying with GPS cannot help but feel armchair mapping is 'cheating' in some way, and the accuracy debate is used to a certain extent to expose 'cheaters'.

It is understandable to an extent, but data is improved over time. With the help of local mappers. Suggesting that because your source MAY have a tiny amount of error in it, then you shouldn't add it is retrograde and holds everything back. Surveying on foot probably has about the same error rate.

Hide this comment

Comment from chriscf on 19 July 2012 at 03:47

You seem to misunderstand what OSM is about. You also seem to have misunderstood my point. I'm not saying that tracing is wrong, I'm saying that tracing from StreetView and then adding the name from Locator is wrong. We're not a dumping ground for any and all data you happen to find. The idea is that people use Locator in addition to, not in lieu of, local knowledge. The purpose of the comparison tools is to tell us where we may be missing something, and to enable us to tell the OS when they get things wrong. I've seen some names in Locator that were patently gibberish. I've seen others that were obvious typos that were actually transferred onto the map. These are actively harmful to the project, not least because we have no way of finding them. FWIW, I can tell you from experience that when it comes to Locator, "MAY have a tiny amount of error in it" is a gross understatement - there are some 7000 errors in Locator that we know about (where someome has seen the name is wrong, and then correctly added a not:name tag to flag it), and surely many more that we don't know about.

Ultimately, before they go near the map, any source of data has to be filtered through a brain. Any process that by-passes this is wrong ab initio.

Hide this comment

Comment from ca_hoot on 19 July 2012 at 11:42

@chriscf I do not misunderstand what OSM is about at all. I fully realise that local knowledge is what makes this map what it is. But your insulting attitude tells anyone reading this exactly what they need to know, 'brainless', 'silly' - have you ever accounted for the fact that maybe the people making errors are just new to OSM and not just 'brainless' as you seem to think. These people should be encouraged not driven away by your little boys club attitude.

Surely the future of OSM is not 'mappers' like us at all. It is people that edit the street they live on or their route to work. They are the people that have real local knowledge. Not you driving through once in your car.

I'm not quite sure who made you the arbiter of what OSM is, but frankly when it comes to misunderstanding I think you are deliberately misunderstanding me, I have said about 5 times now that all care and consideration should be given to other peoples work. In that situation the errors you complain about are rare.

I simply do not accept the fact that using locator to add otherwise nameless roads is wrong. Lots of StreetView names are wrong too are we to discount that? Bing images may be misaligned and out of date. Discount that too? The shoreline might change, don't ever trace that? Streets are renamed, business change, even house numbers change.

There is no perfect map. When we have 10 mappers in every town in the world then maybe your objectives are realistic... but in the meantime, I think using the hundreds of years worth of work the OS have done seems a reasonable stop gap.

Hide this comment

Comment from chriscf on 22 July 2012 at 16:02

have you ever accounted for the fact that maybe the people making errors are just new to OSM and not just 'brainless' as you seem to think

Goodness, some people really are stupid. I "seem to think" because I actually looked to see who was adding those names to make sure they weren't new.

I simply do not accept the fact that using locator to add otherwise nameless roads is wrong.

Whatever. You're free to not accept that water is wet if that suits you. Using Locator to add a name is wrong - end of story. Here, we prefer correct over complete. We'd rather someone add hundreds of names that are all right rather than thousands that could be wrong.

Lots of StreetView names are wrong too are we to discount that?

Yes. If I want a clone of StreetView, I'll use StreetView. StreetView is decent enough for geometry if you have no alternative, but it should not be used for names (which will be duplicated in Locator anyway). There's no margin when it comes to names. Geometry can be improved, but a name is either correct or not correct.

Bing images may be misaligned and out of date. Discount that too?

Misaligned and out of date are not a problem if you have enough local knowledge to recognise it. Misaligned is also not a problem if there's enough data in the area to realign it. A bigger problem is that the imagery often isn't rectified - we've seen an example of this in the diary recently, and I'm aware of at least one part of the world where they actively avoid tracing from Bing for this very reason.

The shoreline might change, don't ever trace that?

Correct. The shoreline visible in an aerial image taken at a non-specific point in the tidal cycle is meaningless, and anyone who is tracing it away from sheer cliff faces should be taken out and shot. That's one place where the OS data comes in useful. Unfortunately, some people have done silly things such as tracing it from StreetView instead of importing the vectors from BoundaryLine or VectorMap. These are useful imports because we've got no other way of obtaining these.

Streets are renamed, business change, even house numbers change.

Now you're just being facetious. This is an area where we have substantially outperformed the traditional mapping agencies.

but in the meantime, I think using the hundreds of years worth of work the OS have done seems a reasonable stop gap.

That's just incredibly lazy. Again, we're not looking for complete data, we're looking for correct data. We're in no rush. We have all the time in the world to "finish" our task, so a quick fix of importing a bunch of data that we could reasonably collect through other means doesn't really help us much. OS data is a supplement to user-contributed data, not a replacement for it. If you find this "insulting", then perhaps OSM really isn't for you.

Hide this comment

Comment from ca_hoot on 23 July 2012 at 07:52

It is impossible to have a constructive discussion with someone as arrogant as yourself Chris. There is no need for all the name calling, or telling people to leave for that matter, that is what I find insulting, not the actual content of what you say.

Your opinion is exactly that, not that of OSM or the people that use it. So don't refer to yourself as 'we'. I have never even slightly suggested that OS data should replace anything - I suggested it should be used where there is nothing.

In terms of this thread, I'm not posting here any more as it has turned into a typical internet debate. I have far better things to do with my time than argue the toss with you.

Hide this comment

Leave a comment

Parsed with Markdown

  • Headings

    # Heading
    ## Subheading

  • Unordered list

    * First item
    * Second item

  • Ordered list

    1. First item
    2. Second item

  • Link

  • Image

    ![Alt text](URL)

Login to leave a comment