Build the EIN -> self-description table

Joins the Form 990 Part III wide table (see dev/build-990-text-wide.R) to the BMF's classification text, one row per EIN. Where an EIN filed more than one return for the year — 8,571 of 544,917 do, from amendments and short periods — the latest tax period wins, then the longest text.

Usage

pf_build_org_text(wide, bmf = NULL, ntee = NULL, cache = NULL)

Arguments

wide

Path to F9-P03-TEXT-WIDE-*.rds/.csv, or the table itself.

bmf

The processed BMF (path or data frame), for the NTEE and activity text. Optional: omit for 990-text-only.

ntee

Output of pf_ntee_text(). Optional.

cache

Path to cache the result as .rds. NULL to skip.

Value

A data.table keyed by ein with ntee_text, activity_text, mission, prog_all, all_text, has_mission, has_program, n_programs, plus a token_idf attribute computed over the mission corpus.