Deterministic Matching
This version of the Texas A&M Geoservices Geocoder performs strictly
deterministic matching, i.e., probabilistic matching is not attempted.
This means that if an exact match is not found in
a particular reference dataset, no match is returned for that reference data layer.
Attribute Relaxation
This option directs the geocoder to try alternative versions of the input data
in the case when an exact match can not be found. In particular, attributes of the
input address are removed from the query, first one at a time and then in combination
with each other. Attribute relaxation is performed (if the option is selected)
on the following address attributes:
|
Rank
|
Attribute
|
| 1) | Street predirectional |
| 2) | Street postdirectional |
| 3) | Street suffix |
| 4) | City |
| 5) | Zip |
An example of the first few iterations it will try are listed in the next table:
| Number |
Pre |
Name |
Suffix |
Post |
Zip |
City |
State |
| 3620 | S | Vermont | Ave | N | 90089 | Los Angeles | Ca |
| 3620 | | Vermont | Ave | N | 90089 | Los Angeles | Ca |
| 3620 | S | Vermont | Ave | | 90089 | Los Angeles | Ca |
| 3620 | S | Vermont | | N | 90089 | Los Angeles | Ca |
| 3620 | | Vermont | Ave | | 90089 | Los Angeles | Ca |
| 3620 | | Vermont | | N | 90089 | Los Angeles | Ca |
| 3620 | S | Vermont | | | 90089 | Los Angeles | Ca |
| 3620 | | Vermont | | | 90089 | Los Angeles | Ca |
Substring Matching
This option directs the geocoder to use substring matching techniques to test
for matches in the database. Using this approach increases the likelyhood of
finding a match if the input data or reference data are incomplete (the recall
is increased). However, using this also increases the chances that wrong results
are returned (the precision is descreased).
The following table shows examples of this strategy:
| Query |
Reference Feature |
Match |
| Vermont |
Vermont |
yes |
| Verm |
Vermont |
yes |
| Mont |
Vermont |
yes |
Soundex Matching
This option directs the geocoder to use soundex matching techniques to test
for matches in the database. Using this approach increases the likelyhood of
finding a match if the input data or reference data have minor misspellings (the recall
is increased). However, using this also increases the chances that wrong results
are returned (the precision is descreased).
The following table shows examples of this strategy:
| Query |
Query Soundex |
Reference Feature |
Reference Feature Soundex |
Match |
| Vermont |
V655 |
Vermont |
V655 |
yes |
| Vermond |
V655 |
Vermont |
V655 |
yes |
| Varnend |
V655 |
Vermont |
V655 |
yes |
-
Parcel centroid
-
A exact match was found to a parcel and its centroid is returned as output
-
Street segment
-
A match was found to the street segment and the address range
associated with the segment was used
to interpolate a point to return as output
-
ZCTA
-
A match was found to the ZIP portion of the address and its centroid was
returned as output
-
City
-
A match was found to the city portion of the address and its centroid was
returned as output
-
County subregion
-
A match was found to the city portion of the address
in the county subregion reference data set and its centroid was
returned as output
-
County
-
A match was found to the city portion of the address
in the county reference data set and its centroid was
returned as output
-
Unmatchable
-
A match could not be found for the input
| Unknown |
| GPS |
| BuildingCentroid |
| Building |
| BuildingDoor |
| Parcel |
| StreetSegment |
| StreetIntersection |
| StreetCentroid |
| USPSZipPlus5 |
| USPSZipPlus4 |
| USPSZipPlus3 |
| USPSZipPlus2 |
| USPSZipPlus1 |
| USPSZip |
| ZCTAPlus5 |
| ZCTAPlus4 |
| ZCTAPlus3 |
| ZCTAPlus2 |
| ZCTAPlus1 |
| ZCTA |
| City |
| ConsolidatedCity |
| MinorCivilDivision |
| CountySubRegion |
| County |
| State |
| Country |
| Unmatchable |
the Texas A&M Geoservices Geocoder allows the user to choose if they want the "best" geocode
returned for an address to be chosen dynamically based on an accuracy
metric calculated by the geocoder, or statically always in the same order.
Details about each of the available methods can be found in the following publication.
Goldberg, D. W., Cockburn, M. G. (2010). Improving geocode accuracy with candidate selection criteria. Transactions in GIS. Vol. 14 (S1), pp. 129-146.
| FeatureClassBased |
| UncertaintySingleFeatureArea |
| UncertaintyMultiFeatureGraviational |
| UncertaintyMultiFeatureTopological |
The uncertainty hierarchy directs the geocoder to choose the geocode
with the lowest uncertainty
as the resulting "best geocode" that should be returned for an address.
This option will slow down the processing of your records.
When this
option is not selected, the "best geocode" will be chosen based on the first
geocode that matches in the following table (a variant of the NAACCR Hierarchy):
| Exact parcel centroid |
| Nearest parcel centroid |
| Uniform lot interpolation |
| Address range interpolation |
| ZIP code centroid |
| City centroid |
| County subdivision centroid |
| County centroid |
| State centroid |
| Country centroid |
-
StreetAddress
-
The matched input data was a postal street address
Example: 3620 South Vermont Ave, Los Angeles, CA 90089-0255
-
PostOfficeBox
-
The matched input data was a Post Office Box address
Example: PO Box 0255, Los Angeles, CA 90089-0255
-
RuralRoute
-
The matched input data was a Rural Route address
Example: RR 13 Box 2, Los Angeles, CA 90089-0255
-
StarRoute
-
The matched input data was a Star Route address
Example: Star Route 13 Box 2, Los Angeles, CA 90089-0255
-
HighwayContractRoute
-
The matched input data was a Highway Contract Route address
Example: HC 13 Box 2, Los Angeles, CA 90089-0255
-
Intersection
-
The matched input data was an intersection of two or more streets
Example: 36th and Vermont, Los Angeles, CA
-
NamedPlace
-
The matched input data was a named place
Example: USC GIS Research Laboratory, Los Angeles, CA
-
RelativeDirection
-
The matched input data was a relative direction
Example: 1 mile south of downtown Los Angeles
-
Unmatchable
-
A match could not be found for the input
| Unmatchable |
| Unknown |
| StreetAddress |
| PostOfficeBox |
| RuralRoute |
| StarRoute |
| HighwayContractRoute |
| Intersection |
| NamedPlace |
| RelativeDirection |
| USPSZIP |
| City |
| State |
-
Exact
-
The input data exactly matched a feature in the reference data source:
| input |
3620 S Vermont Ave, Los Angeles, CA 90089-0255 |
| reference |
3620 S Vermont Ave, Los Angeles, CA 90089-0255 |
-
Relaxed
-
One or more of the input data attributes had to be removed to find a match in the reference data source:
| input |
3620 Vermont Ave, Los Angeles, CA 90089-0255 |
| reference |
3620 S Vermont Ave, Los Angeles, CA 90089-0255 |
- South is missing from input and is present in the reference
For more details, see the section on Address Relaxation above.
-
Soundex
-
One or more of the input data attributes had to be matched with soundex to find a match in the reference data source:
| input |
3620 S Vermonnt Ave, Los Angeles, CA 90089-0255 |
| reference |
3620 S Vermont Ave, Los Angeles, CA 90089-0255 |
- Soundex("Vermonnt") = V655 = Soundex("Vermont")
For more details, see the section on Soundex Matching above.
| NoMatch |
| Exact |
| Relaxed |
| Substring |
| Soundex |
| Composite |
| Nearby |
| Unknown |
-
Success
-
A exact match was found
-
Unmatchable
-
A match could not be found for the input
| Unknown |
| Success |
| Ambiguous |
| BrokenTie |
| Composite |
| Nearby |
| LessThanMinimumScore |
| InvalidFeature |
| NullFeature |
| Unmatchable |
| ExceptionOccurred |
| Unknown |
| RevertToHierarchy |
| FlipACoin |
| DynamicFeatureComposition |
| RegionalCharacteristics |
| ReturnAll |
| ChooseFirstOne |
