README.Rmd

---
output:
  md_document:
    variant: gfm
bibliography: refs.bib
---

<!-- README.md is generated from README.Rmd. Please edit that file -->

```{r, echo = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-"
)
```

# dblinkR: An R interface for dblink

## Overview
`dblinkR` is an R interface for 
[`dblink`](https://github.com/cleanzr/dblink)—an Apache Spark package for 
performing unsupervised entity resolution. 
It implements a generative Bayesian model for entity resolution called `blink` 
[@steorts_entity_2015], with extensions proposed in [@marchant_d-blink_2021].
Unlike many entity resolution methods, `dblink` approximates the full posterior 
distribution over the _linkage structure_.
This facilitates propagation of uncertainty to post-entity resolution 
analysis, and provides a framework for answering probabilistic queries about 
entity membership.

## Installation
`dblinkR` is not currently available on CRAN. 
The latest development version can be installed from source using `devtools` 
as follows:

```{r, eval=FALSE}
library(devtools)
install_github("ngmarchant/dblinkR")
```

### Dependencies
`dblinkR` depends heavily on the `sparklyr` R interface for Apache Spark. 
Please refer to the `sparklyr` [website](https://spark.rstudio.com/) for 
information about connecting to a Spark deployment. 

`dblinkR` currently supports Spark releases in the 2.3.x series and 2.4.x 
series. Spark releases prior to 2.3.x are not supported.

## Example
The  [RLdata500 vignette](vignettes/RLdata500.Rmd) demonstrates how to 
use `dblinkR` to perform entity resolution for a small synthetic 
data set. This example is small enough to run on a laptop (Spark 
cluster not required).

## Licence
GPL-3

## References