Skip to content

Getting Started


Splink supports python 3.8+.

To obtain the latest released version of Splink you can install from PyPI using pip:

pip install splink

or if you prefer, you can instead install Splink using conda:

conda install -c conda-forge splink
Backend Specific Installs

Backend Specific Installs

From Splink v3.9.7, packages required by specific Splink backends can be optionally installed by adding the [<backend>] suffix to the end of your pip install.

Note that SQLite and DuckDB come packaged with Splink and do not need to be optionally installed.

The following backends are supported:

pip install 'splink[spark]'
pip install 'splink[athena]'
pip install 'splink[postgres]'
DuckDB-less Installation

DuckDB-less Installation

Should you be unable to install DuckDB to your local machine, you can still run Splink without the DuckDB dependency using a small workaround.

To start, install the latest released version of Splink from PyPI without any dependencies using:

pip install splink --no-deps

Then, to install the remaining requirements, download the following requirements.txt from our github repository using:


# Download the file from GitHub using curl
curl -o "$output_file" "$github_url"

Or, if you're either unable to download it directly from github or you'd rather create the file manually, simply:

  1. Create a file called splink_requirements.txt
  2. Copy and paste the contents from our duckdbless requirements file into your file.

Finally, run the following command within your virtual environment to install the remaining Splink dependencies:

pip install -r splink_requirements.txt

🚀 Quickstart

To get a basic Splink model up and running, use the following code. It demonstrates how to:

  1. Estimate the parameters of a deduplication model
  2. Use the parameter estimates to identify duplicate records
  3. Use clustering to generate an estimated unique person ID.

For more detailed tutorial, please see section below.

Simple Splink Model Example
from splink.duckdb.linker import DuckDBLinker
import splink.duckdb.comparison_library as cl
import splink.duckdb.comparison_template_library as ctl
from splink.duckdb.blocking_rule_library import block_on
from splink.datasets import splink_datasets

df = splink_datasets.fake_1000

settings = {
    "link_type": "dedupe_only",
    "blocking_rules_to_generate_predictions": [
    "comparisons": [
        ctl.date_comparison("dob", cast_strings_to_date=True),
        cl.exact_match("city", term_frequency_adjustments=True),
        ctl.email_comparison("email", include_username_fuzzy_level=False),

linker = DuckDBLinker(df, settings)

blocking_rule_for_training = block_on(["first_name", "surname"])

linker.estimate_parameters_using_expectation_maximisation(blocking_rule_for_training, estimate_without_term_frequencies=True)

blocking_rule_for_training = block_on("substr(dob, 1, 4)")  # block on year
linker.estimate_parameters_using_expectation_maximisation(blocking_rule_for_training, estimate_without_term_frequencies=True)

pairwise_predictions = linker.predict()

clusters = linker.cluster_pairwise_predictions_at_threshold(pairwise_predictions, 0.95)

🔗 Tutorials

You can learn more about Splink in the step-by-step tutorial.


Example Notebooks

You can see end-to-end example of several use cases in the example notebooks, or by clicking the following Binder link:


You can see all of the interactive charts provided in Splink by checking out the Charts Gallery.