Normalization, the name-frequency table, the token-IDF table, and the
blocking index are all functions of the reference alone, so they are computed
once here and reused across every pass, batch, and experiment. This is
preprocessing and is excluded from the match timings reported by
pf_match_grants().
Usage
pf_build_reference(bmf, active_col = NULL)
Arguments
- bmf
The raw processed BMF data frame (or a path to its CSV).
- active_col
Optional column flagging BMF-active EINs (for the unified
BMF). Omit for the active-only file.
Value
A list with the normalized reference, name_freq, token_idf, and
ref_index, plus the elapsed seconds each stage took.