RERBOT
R package · xinminchu/erbot · MIT

Entity resolution,
sand in, gold out.

ERBOT is a unified R pipeline for deduplication and record linkage — nine stages from raw CSV to evaluated clusters, with the evaluation discipline most toolkits skip.

Try the Studio →Read the docs
# install
remotes::install_github("xinminchu/erbot")

# one call, nine stages
res <- er_run(df, truth = truth_tbl)
res$metrics  # ARI, B-cubed, AMI
199
regression tests, zero warnings
9
pipeline stages, one function call
11
supervised classifier families
0
truth leakage in default paths
Nine stages, one call
Load → diagnose → block → similarity → combine → weights → cluster → merge → evaluate.
How it works →
Honest by default
Truth is evaluation-only. Entity-disjoint splits. No silent misalignment.
The methodology →
Documented & tested
Every stage is a tested, documented function. R package on GitHub, MIT licensed.
Read the docs →

Have messy data? Run it.

Upload a CSV in the Studio — no install, no signup — and get clusters, metrics, and the equivalent R code.

Open the Studio →