A first-pass thematic map of the correspondence — twelve subjects, ranked and traced by decade. The scaffold for the book's chapters and the advice-mapping ahead.
Ranked by primary topic — each letter's dominant theme. Torah & scholarship lead, then chinuch, blessings and Eretz Yisroel. Many letters also carry secondary tags (in the data file).
Each row is shaded against its own peak decade, so a theme's shape shows through the 1950s bulge. Note the late tilt of blessings & telegrams and shidduchim toward the 1970s.
Cell = letters whose primary topic is this row, in that decade · shade = share of the row's own maximum.
Tagged first by a transparent Hebrew keyword lexicon (strong terms tag alone; generic terms need support), multi-label with a ranked primary, then extended by a semantic reading pass that placed 1,051 of the residual untagged letters. Coverage is now 90%; the remaining 10% untagged are mostly short notes, administrative apparatus, and the OCR volumes. Science & Torah (16) stays hand-restricted — its lexicon over-fires. Reusable outputs: igros_topics.jsonl (per-letter tags; semantic rows carry src:"semantic") and igros_topic_stats.json.