flap-lite 0.6.17

An open-source tool for linking free-text addresses to UPRN

FLAP

FLAP is an open-source tool for linking free-text addresses to Ordinance Survey Unique Property Reference Number (OS UPRN). You need to have a licence of OS UPRN and download the address premium product to use FLAP FLAP can be used at scale with a few lines of syntax.

Quick start of FLAP tool

Installation

We recommend you to create a virtual environment with venv.

python3 -m venv [YOUR_PATH]/flap_lite
source [YOUR_PATH]/flap_lite/bin/activate

Install with pip:

pip install --upgrade flap-lite

For now, please contact the developer for downloading the trained model. Copy the model to [YOUR_PATH]/flap_lite/lib/python3.9/site-packages/flap/model/

cp [PATH_TO_MODEL_FILE] [YOUR_PATH]/flap_lite/lib/python3.9/site-packages/flap/model/

Quick Start

Building the database

Use flap.create_database for building the database.

from flap import create_database

create_database(db_path=[PATH_FOR_THE_DB], raw_db_file=[PATH_TO_DB_ZIP])

Matching

Use flap.match for matching address to database

from flap import match

input_csv = '[PATH_TO_INPUT_CSV_FILE]'
db_path = '[PATH_TO_THE_DB]'

results = match(
    input_csv=input_csv,
    db_path=db_path
    )

Matching results will be saved to [$pwd]/output.csv by default. By default, FLAP uses all available CPUs and process the addresses in batches of 10,000.

Some useful options are:

• batch_size for number of addresses in each batch
• max_workers for CPU cores used
• in_memory_db for if in-memory SQLite is used

How does it work?

Briefly, FLAP parses the structured parts of addresses (e.g. POSTCODE "AB12 3CD"). And all the deterministic parts (e.g. numbers "111", letters "A")

An SQL query is made based on the parsed fields to narrow down to a few rows in the database.

select * from indexed where POSTCODE='AB12 3CD'

Features are generated:

• For the deterministic parts: pairwise comparison to see if equal
• Linear assignment alignment for the textual parts
• For postcode: comparison to see if parts are equal

A trained Random Forest Classifier predict a score based on the generated feature. The address with best score is deemed as a match.

The above is a simplified description.

Coming soon

• Command Line Interface
• More documentation
• Dummy database for trying it out with an example notebook