Artificial Intelligence for Better Geocoding
There are two parts to geocoding: a good address and a good dataset against which to geocode it.
A researcher at Curtin University of Technology in Perth, Australia is working on the first part. In particular, taking out some of the human error in keying in addresses. The software, a protoype at this point, is called IntelliGeoLocator. It’s developer is Matthew Hutchinson.
The head of the spatial sciences department at Curtin University, Bert Veenendaal, says about 70 per cent to 90 per cent of addresses can be geocoded but the remainder might be more complicated due to missing information and wrong addresses.
“The software understands how addresses work rather than matching it to lists of known addresses, which is what current geocoders do,” Veenendaal says.
This would be, in my eyes, a “real time” type of address cleaning, as opposed to the batch tools offered by Pitney Bowes and others.
