[HOT] Broken (multi)polygon cleanup

Jochen Topf jochen at remote.org
Fri Mar 17 12:16:26 UTC 2017


I am sure there are many other problems, but can we please keep to the
issue at hand in this thread? I am not trying to solve all problems, I
am trying to solve one particular set of problem related to

And I didn't pick that particular problem at random. This is something
that affects everybody. The short of it is: Due to upcoming software
changes, some (multi)polygons that are rendered now will disappear from
the map or look different in the future. Again, you can find all the
details about the problems and of what we are trying to achieve at
http://area.jochentopf.com/ .

We are having great success working on these problems. You can see the
stats at http://area.jochentopf.com/stats/ . This doesn't mean that the
approach will be the best one for other problems, and it doesn't mean we
can't do even better than we are doing until now with this problem.

Btw: I am deliberately talking about "(multi)polygons". This is not
something that only affects multipolygon relations, but all polygons,
even the ones created from closed ways. They are simpler than mp
relations, but that doesn't mean they can't have problems.


On Thu, Mar 16, 2017 at 05:23:47PM -0400, john whelan wrote:
> Date: Thu, 16 Mar 2017 17:23:47 -0400
> From: john whelan <jwhelan0112 at gmail.com>
> To: Jochen Topf <jochen at remote.org>
> Cc: hot at openstreetmap.org, Blake Girardot HOT/OSM
>  <blake.girardot at hotosm.org>
> Subject: Re: [HOT] Broken (multi)polygon cleanup
> This is similar to my pulling in Africa a country at a time and using JOSM
> to check for errors.
> What I have noticed is one mapper will misstag then others will follow.  So
> sanitising the area or checking it for errors before the new HOT mappers
> start would probably improve the overall data quality as they tend to
> follow what is already mapped.
> Untagged ways by one mapper tend to cluster.  I've seen a 100 buildings in
> one small area by the same mapper left as area=yes.  That's quick and easy
> to fix.
> Recently I came across 42 highways left untagged by the same mapper in the
> same area.  The mapper had more than 2,000 edits so wasn't inexperienced.
> Crossing highways often I'll see a town where the highway residentials are
> laid out in a grid but the junctions aren't nodes.
> Another nice recent one was a batch of settlements that had been tagged
> highway=unclassified rather than landuse=residential.  Searching for
> highways above unclassified and crossing ways picked those out quite
> quickly.
> These types of errors are handled more quickly directly in JOSM than
> involving maproulette.
> Even the some other errors can be loaded up in the JOSM  to-do plugin.
> I've been cleaning up in Nigeria for a while now.  Making repeated passes
> over the same areas over time.  I'm seeing fewer errors so my suspicion is
> the improved learnOSM, tutorials and simplifying what we ask mappers to map
> is paying off.  Validating new mappers work within 48 hours does pay off in
> data quality terms.  Giving feedback more than two weeks later is usually a
> waste of time based on my own experience and Martin's report but I don't
> think we have the validators to do the validation within the magic 24/48
> hours.
> Broken multipolyons do take a level of expertise to resolve.  I usually
> avoid them.  However new mappers will fearlessly tread ahead and it is to
> be expected they will break some.
> I don't think learnOSM goes into detail about them or about relationships.
> Should they?  That's another question.
> For some errors we need boots on the ground.  Especially when it comes to
> urban areas with limited access highways such as dual carriageways​ and a
> highway approaching them can you turn in either direction?
> Often the most motivated people to fix the errors are the locals who will
> be using the map.  Trouble is we first have to get something in place that
> is useful to get them motivated.
> So yes HOT maperthons do have a reputation of the blind leading the blind,
> ie some are led by inexperienced mappers but we are putting effort into
> reducing the number of errors.  The training group has put together a very
> good detailed set of instructions on how best to run a maperthon and if you
> can identify a problem area HOT will listen and try to address the issues.
> Cheerio John
> On 16 Mar 2017 3:49 pm, "Jochen Topf" <jochen at remote.org> wrote:
> > Hi Blake,
> >
> > On Wed, Mar 15, 2017 at 05:29:04PM +0100, Blake Girardot HOT/OSM wrote:
> > > Can you send me something that would help me see those 1600 examples
> > > you found on OSM so I can see what might be going on and how to
> > > improve things in HOT/MM workflow to reduce them?
> >
> > Those 1600 are some very specific errors I am currently looking at, but
> > only have those on my local machine currently. But for the wider picture
> > you can use the OSM Inspectors "area" view to see the scope of the problem.
> >
> > Just go to any place where HOT/MM was active. Here is an example:
> > http://tools.geofabrik.de/osmi/?view=areas&lon=33.65927&
> > lat=-2.07600&zoom=8
> >
> > My current workflow is as follows: I choose one particular well-defined
> > problem and extract the data from OSM that shows this problem. Then I
> > put the data into Maproulette and document the problem and how to fix
> > it. Some problems are easier, some harder to fix. Mappers can choose
> > what problems to work on based on their skill level. If somebody
> > doesn't feel comfortable with one of the harder problems, there are
> > simpler ones they can work on. I also have some more general
> > descriptions for the experts who want to go exploring/fixing on their
> > own. Once the challenges are done, I choose a new problem and start the
> > process from the beginning.
> >
> > Here are the current challenges: http://area.jochentopf.com/fixing.html
> >
> > Usually I am splitting up larger challenges into regions, for instance
> > by continent or country. I could also do this by HOT activation area or
> > so if that makes sense for HOT mappers. Splitting up, often makes the
> > problems easier to fix, because problems inside an area are often very
> > similar, but different to problems in other areas due to mapping
> > priorities and methods. For instance, some regions have lots of problems
> > with buildings, others with landuse areas, etc. Concentrating on one
> > very well-defined problem at a time makes all this work simpler.
> >
> > Jochen
> > --
> > Jochen Topf  jochen at remote.org  https://www.jochentopf.com/
> > +49-351-31778688
> >
> > _______________________________________________
> > HOT mailing list
> > HOT at openstreetmap.org
> > https://lists.openstreetmap.org/listinfo/hot
> >

Jochen Topf  jochen at remote.org  https://www.jochentopf.com/  +49-351-31778688

More information about the HOT mailing list