Readme

The hyrule package contains functions to facilitate record linkage (i.e., entity resolution) using machine learning. In the example_workflow subdirectory, there is a fully worked out linkage pipeline that demonstrates how hyrule, along with a few other packages, can be combined/used to conduct record linkage.

There are three main vignettes/examples:

1. Record Linkage Pipeline: A mostly comprehensive example of a targets analysis pipeline for machine learning record linkage. The _targets.qmd file uses targets-flavored markdown and can be edited (or stripped for parts) into a record linkage pipeline using “real” data. To “recreate” the document users must first render the .qmd file, execute the pipeline via tar_make(), and re-render the .qmd file. The double render is required to populate the parts of the document that rely on completed results.
2. Evaluating results: Opinions and ideas on how to tell if a record linkage is any good
3. Generating new training data: A description of some ways to generate effective training data, including a review of the hyrule::matchmaker() function/shiny app.