[Talk-GB] OSM Analysis updated with new OS Locator data and a review of progress to date

[Peter Miller]

On 8 June 2011 14:18, Ed Avis <eda at waniasset.com> wrote:
> Steve Doerr <doerr.stephen at ...> writes:
>
>>I wonder if the good folks at ITO could devise a way to analyse the
>>not:name tags in the database and see whether any of them are now
>>redundant? In other words, are the OS correcting any of the mistakes we
>>appear to have identified?
>
> It would be cool to see a comparison the other way round: testing the OS data
> for accuracy using OSM as a reference.  In inner London I think there are about
> 5% of names missing from OS - mostly semi-private drives or estates, but
> nonetheless signposted and addressable - so I think they would score no higher
> than 95%.
>
> (OS Street View is a bit better, I'd say that only about 2% of roads that exist
> are missing from it, and the 'false positive rate' of Street View showing a road
> where nothing is on the ground is almost nil.  It's not as easy to do automated
> comparisons however.  These numbers are totally off the top of my head and apply
> to London only.)

I do agree that it may now be interesting to include two new columns:

1) A list of not:names that orginated from OS Locator but where OS
Locator does not currently contain that error. The challenge is that
not all not:name entries in OSM will have originated from error in OS
Locator; they could contain details of errors from other sources, such
as Navteq or TeleAtlas or elsewhere. The practical approach may be to
just publish the differences and not worry about the original source
but include text from the not:name:note field which can provide any
supporting information about where the error came from (such as "duff
data in TeleAtlas Oct10", or "OS Locator June10").

2) A list of street names which are in OSM but which are not in OS
Locator could be a good publicity tool for OSM and a good new source
of errors for elements of a way (for example where a short section of
a street associated with a bridge but the other way had a typo in
OSM). I guess that needs would ideally have its own rendering layer?
We might start with just a list on district page with no rendering and
we come back to rendering at a later point. If others wished to create
rendering now then that would be great!

Finally. Might it be useful for us to accommodate have multiple
not:name entries associated with a single road? For example where a
single street has multiple different duff names from one or more
different sources, ie OS Locator and Navteq both have different wrong
names. Should we accommodate  'not:name_xxx' where xxx can be any
text?


Regards,


Peter

>
> --
> Ed Avis <eda at waniasset.com>
>
>
>
>
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb
>