JTaber wrote:Which is why I use batchgeocode.com, which uses Yahoo Maps, strictly as a backup. Specifically, if it can't find a street, it often places the point in the middle of the ZIP Code; if the ZIP Code is off (as it sometimes is, since I don't have time to check all of them from what the wards have) it can get confused. If it places the point at least in the right ward, I'll usually take what I can get though.
That is a property of batchgeocode.com, not of Yahoo. What the batchgeocode.com web page does (client-side) is send repeated calls to the Yahoo geocoding API. Yahoo actually returns two additional data attributes for each address, describing the quality of the geocoding transaction. But the implementer of batchgeocode.com -- which is otherwise nicely done -- chose not to do anything with those attributes. It just accepts the coding to a zip centroid without so much as a warning.
I found this unacceptable. So I wrote my own VBScript that talks to the Yahoo API directly, and retrieves the addresses along with the precision attributes. Then I can deal with the exceptions offline before plotting the map.
Most often the root cause of a bad geocoding transaction is poorly formatted input data in MLS. Garbage In, Garbage Out.
My script does do some minimal preprocessing of the input data to scrub some common style errors in addresses. I plan to experiment with preprocssing the addresses further through calls to a Postal Service API, but that may be gilding the lily.
I have found the quality of Yahoo's lat/lon geocoding to be quite good for my urban location. Of about 570 records I submitted for our huge ward, all but about 7 errors were due to bad address data. On these 7 records, Yahoo's geocoding database could not find the correctly formatted addresses, but I was able to geocode them manually on other sites such as Google. So I ended up with 100 percent of the geocodable records successfully geocoded. I am working on a way to store such exceptions locally so the work doesn't have to be repeated for the next update..
BTW, my own solution --a work in progress -- uses .Yahoo's geocoder, but the output is KML for display in Google Earth. ( I also have a precise ward boundary layer that I produced in commercial GIS software, but that's a separate piece. Easily done with the right tools.)
The smarts offered in Google Earth's interface, which runs as a local application, surpass those of a purely web-based viewer.
The next step I see is producing Google Earth hierachies of HT/VT routes for the appropiate leaders to use. Beyond just visualizing, they should be able to do
interacive what-if geographic editing of routes inside Google Earth, then export the results. (The only thing missing would be integration with MLS
p.s. I'm new here. Thanks to the work of those implementers upthread, whose ideas I have adapted.