In Ironholds/poster: Street Address Normalisation and Parsing

Using the poster package

libpostal is a postal address parser written in C - and poster is a wrapper around it, allowing you to parse and normalise addresses from R.

Installation

Installing poster is pretty easy - just run install_github or install.packages as appropriate - as long as you've got libpostal on your system. Installing libpostal can be done with the following; on Linux:

sudo apt-get install curl libsnappy-dev autoconf automake libtool pkg-config
git clone https://github.com/openvenues/libpostal
cd libpostal
./bootstrap.sh
./configure --datadir=[...some dir with a few GB of space...]
make
sudo make install
sudo ldconfig

On Macs:

sudo brew install snappy autoconf automake libtool pkg-config
git clone https://github.com/openvenues/libpostal
cd libpostal
./bootstrap.sh
./configure --datadir=[...some dir with a few GB of space...]
make
sudo make install

libpostal doesn't run on Windows - and correspondingly poster won't.

Features

The fastest and simplest feature of poster is an address normaliser, which does what it says on the tin: uses statistics to try and turn the awkward, vague, inconsistent addresses human beings put into forms into a standard format. This can be handled with the normalise_addr function:

normalise_addr("Quatre-vignt-douze Ave des Champs-Élysées")
[1] "4-vignt-12 avenue des champs-elysees"

More importantly, poster can also parse addresses, standardising it and splitting it up into everything from the house name or number to the postal code and country:

str(parse_addr("781 Franklin Ave Crown Heights Brooklyn NYC NY 11216 USA"))
'data.frame':   1 obs. of  10 variables:
 $ house         : chr NA
 $ house_number  : chr "781"
 $ road          : chr "franklin ave"
 $ suburb        : chr "crown heights"
 $ city_district : chr "brooklyn"
 $ city          : chr "nyc"
 $ state_district: chr NA
 $ state         : chr "ny"
 $ postal_code   : chr NA
 $ country       : chr "usa"

As well as parse the entire address, you can also extract individual elements: there are functions with names matching the columns in parse_addr's output (house, house_number, country) that extract those individual elements.