Street Address Validation

Introduction

In collaboration with the Treasurer/Tax Collector we are making a restful street address geocoding service available.
This page describes the service.
We mostly follow the OGC geocoding standard which is described here.

The Request

Here is the DEV URL with an example address of "100 Main St":

http://10.250.60.189/geocode/streetNetwork/findAddressCandidates?f=json&Address=100 Main St&Zip=94105

The service uses the street segment network only.
These data are carefully maintained by Department of Public Works, and brought into EAS on a nightly basis.
We return a single best candidate and have not implemented a scoring system.
Instead of scoring we return exceptions to help you correct the input data.
The location is based on the linear interpolation of the matching street segment.
We offset by 4 meters (about 12 feet) to the even or odd side like this:

The Response

We strongly prefer JSON (available now) but we can produce XML if you buy us a beer.
The JSON is formatted for readability but your browser may obscure that.
The way to see the nicely formated JSON is to right click on the web page and select "view source".
The Chrome browser has a nice feature here: (You need to copy and paste the text below to the Chrome address bar and hit enter.)

view-source:http://10.250.60.189/geocode/streetNetwork/findAddressCandidates?f=json&Address=169 SACRAMENTO ST&Zip=94114

In any case, the JSON should look something like this:

{
    "inputAddressString": "100 Main St", 
    "inputZipCodeString": "94105", 
    "spatialReference": {
        "wkid": 4326
    }, 
    "foundMatch": true, 
    "foundEasMatch": false, 
    "candidates": [
        {
            "address": "100 MAIN ST", 
            "location": "POINT (-122.3948959209805700 37.7917552676099080)", 
            "score": null, 
            "attributes": {
                "StreetName": "MAIN", 
                "StreetType": "ST"
            }, 
            "addressNumber": "100", 
            "street": {
                "streetNamePrimary": {
                    "base_street_name": "MAIN", 
                    "street_type": "ST"
                }, 
                "streetNameAliases": [], 
                "l_f_add": 101, 
                "l_t_add": 199, 
                "r_f_add": 100, 
                "r_t_add": 198, 
                "seg_cnn": 8628000, 
                "geometry": "MULTILINESTRING ((-122.3949402377882200 37.7918420903234780, -122.3934079394635700 37.7906126704266400))"
            }, 
            "zipCode": "94105", 
            "jurisdiction": "SF MAIN"
        }
    ], 
    "validations": [], 
    "didYouMean": []
}

Results, Performance, Tests

We have tested the DEV implementation against all the Treasurers data with these results:

found street match	found eas match	count	percent
false	false	1444	2.72
true	false	6481	12.24
true	true	44996	85.02
		52921	100

The service processes about 10 addresses per second which should be fine for our purposes. We have done no performance optimization so we can probably make it faster if you have a compelling reason.

We have written unit tests which you can see here.