Address Geocoding - Introducing Geographic Information Systems with ArcGIS - page 576

Making use of addresses requires either personal or institutional knowledge of the location, or requires
a map. To use a map, one must first find the right map—not necessarily an easy task and not one
guaranteed of success if the address is in a sparsely populated area.
Suppose that we have this problem: A municipality wants to determine the spatial distribution of
burglaries. It has records of the addresses of the crimes. One way to do this would be to put a map on a
wall and mark each instance according to the address. Person A might look at a police report and call out
the street address. Person B would look at an alphabetical list of streets and announce the row-column
zone that the street was in (e.g., C-6). Person C could then locate the street on the map and make a guess
at the position of the structure along the street by knowing the house number. In other words, this is not a
particularly simple operation. What would be required to perform the equivalent process with a GIS?
First, we would need some sort of base map showing streets that could be referenced by spatial coordinates.
Next, we would need an extensive database that related locations of structures to the coordinate system
of the base map. Finally, we need some automated way to parse 16 addresses, such as those exemplified in
the preceding list, so that they may be matched against the database. Such a parser must be sophisticated
enough to derive the “true” address from any of a large number of text strings that might represent an
address in varying forms (with variation in blanks, commas, hyphens, letter case, and so on).
TIGER/Line Files
Determining geographic coordinates from addresses is a tall order. Let's begin by restricting ourselves
to the United States and the forms of address used there. You have already met TIGER/Line files,
developed by the Bureau of the Census. If the terms behind the acronym 17 TIGER leave you confused,
consider the predecessor description of such files: DIME, standing for Dual Independent Map Encoding.
The idea behind “dual” is this: Each record in the file contains two geographic references. One is a
range of address; the other is latitude-longitude coordinates. So this sort of file ties together (a) locations
specified by text strings with (b) locations that are specified by geographic coordinates.
Each record in the TIGER file specifies a geographic line (called a chain). For streets in urban and sub
urban areas, the chain usually represents a single block. At the ends of chains are intersections. See
Figure 9-16.
Each record also contains the street name (Ninth, Jenkins, etc) and street type (Lane, Drive, etc.) together
with any prefixes and suffixes (NW, E, and so on). The referenced line has a beginning point and an
ending point. Each such point is defined by a latitude and longitude pair, in the NAD 1983 geographic
(latitude, longitude) coordinate system. Further, at the beginning of each line, two structure numbers
(street address numbers) are specified—one for the structure on the left and one for the structure on the
right. The same is true for the ending point. Finally, if the line curves, its path is prescribed by vertices—
familiar to you from your previous work with linear, vector representations in shapefiles, coverages, and
geodatabase feature classes.
So, you can see that the TIGER database has all the ingredients that are logically necessary to convert
an address into coordinates that will be “spatially close” to the actual coordinates of the structure at
the address. A record in TIGER also contains lots of other information, useful in taking the census and
16 Parse: To breakdown a sequence of letters or numbers into meaningful parts, based on a set of rules.
17 Topologically Integrated Geographic Encoding and Referencing