Hello,
I am trying to compare ThinkGeo Geocoder to other available geocoders. I have been mainly comparing it to MapPoint Web Services (see msdn.microsoft.com/en-us/library/cc534868.aspx and mappoint-css.live.com/CSCV3). Given that, I have a number of questions, especially concerning the analysis of the quality of the results from geocoding.
Here are some statistics that MapPoint Web Service provided as a result of geocoding on a per record level:
MatchCode: msdn.microsoft.com/en-us/library/cc546334.aspx
Good
0
The geocoder found a single match.
Bad
1
The geocoder was unable to return a result.
Imprecise
2
The geocoder was unable to match all of the information and had to geocode up the hierarchy to find a less precise result. For example, the geocoder returns a match code of Imprecise if it cannot match the address fields for the entity and returns a match at the postal code level.
Ambiguous
3
The geocoder returned more than one possible match. For example, the street address for the entity is 128 Main St. and the geocoder returns matches for 128 North Main St. and 128 South Main St. because there is no way to determine which is desired. By default, to the geocoder accepts the first match in the list and stores the associated latitude and longitude coordinates. If you do not want the upload process to accept the first result, select the Reject ambiguous matches check box on the Create a Data Source page or the Replace a Data Source page. If you select this check box, no results are accepted and you can use the Edit Entity ID page to choose the correct match manually.
Disambiguated
4
The match returned by the geocoder has been manually set by the user in the Edit Entity ID page. The Matched Method code is 6 (Manual).
Matched Method: msdn.microsoft.com/en-us/library/cc546335.aspx
Address
0
The geocoder matched the full address, including street number, and used interpolation to return a latitude and longitude coordinate of the address
Street
1
The geocoder matched the street name. The point returned by the geocoder is one end of the street
PostCode
2
The geocoder matched the postal code and returned a latitude and longitude coordinate for the postal code centroid (A centroid is the center of a body’s mass. When the term centroid is applied to a geopolitical area, such as a postal code area or city, the centroid is the center of the population mass of that area rather than the geographic center. Therefore, the centroid for a postal code can be some distance from its geographic center.)
City
3
The geocoder matched the postal code and returned a latitude and longitude coordinate for the city centroid
Subdivision
4
The geocoder matched the state, province, or other subdivision
CountryRegion
5
The geocoder matched the country/region
Manual
6
The results of the geocoder are overriden by a manual fix
Rooftop
7
The geocoder matched the full address and returned a latitude and longitude coordinate of the rooftop at that address
I have two questions in regards to ThinkGeo Geocoder:
- Similarly to MapPoint web services, can geocoder fall back to higher level entites when lowest level match fails (ie, first match on Address, then on Street, then on PostCode and so on). From what I have seen in the samples, it seems that it can fall back; however, not as gracefully: for example, if I enter a street name and city, state (no street number) it falls back to the center of the city - I would expect it to fall back to the street name (I was using maps.thinkgeo.com/web/ to test).
- Can ThinkGeo geocoder report match result codes - as above. Both of the structures above (or similar) are relatively important to understand how well the geocoder has been able to match the actual location: especially if it is run with large data set in a batch.
Thank you