<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<meta content="text/html; charset=ISO-8859-1"
<body bgcolor="#ffffff" text="#000000">
On 15/02/2011 17:23, Ed Avis wrote:
<pre wrap="">Andy Robinson (blackadder-lists <a class="moz-txt-link-rfc2396E" href="mailto:ajrlists@..."><ajrlists@...></a> writes:
<pre wrap="">Great work! How can you tell when you have every postcode and is there
<pre wrap="">some way of checking them against the OS OpenData postcode centroids?
Just by being systematic. If you have Chillly's codepoint postcode layer
sitting over BING its easy in the editor to assign the postcodes as you see
them when adding buildings.
<a class="moz-txt-link-rfc2396E" href="http://www.raggedred.net/codepoint/"><http://www.raggedred.net/codepoint/></a>
I see - so you can use your judgement to work out the area of a postcode based
on its centroid and the streets and buildings nearby.
Since this isn't using any kind of ground survey to check the data, I wonder if
it would be a suitable task for a bot? A program making guesses about postcode
areas might not be any more fallible than a human doing the same task. Of course
it could only be done in areas that had already reached a high standard of
completeness, ideally with buildings traced as well as streets.
Or, perhaps, the robot could make suggestions which a human would then accept or
reject, so that 95% of the area could be covered, with human assistance for the
last few tricky bits.
This just doesn't work. The example I pointed to has shared
semi-detached house which have different postcodes: there is no way
in which any program could work this out, unless it already has
housenumbers which have been mapped using, <i>inter alia</i>, shoe
leather. Adding postcodes to ground surveyed data is relatively
trivial, do not assume that the operation is commutative.<br>