SandX

SandX is the R package that carries ERBOT’s blocking research: collective blocking strategies for entity resolution at scale, evaluated with the discipline the project demands — every experiment ships with a pre-written kill line, and a non-positive result counts as a failure, not a footnote.

What it does

The package implements the blocking methods explored in the ERBOT advisor-feedback track (scenario presets, stratified sampling designs, similarity layers for synthetic benchmarks), with friendly errors and mechanism cards explaining why each success or failure happened.

Status and limitations

SandX is research software under active development. APIs are unstable, large-scale validation is pending, and nothing here has been peer-reviewed. Use it to reproduce experiments, not as a production dependency.