10 Commits
Author SHA1 Message Date
Alex-GilbertandClaude Opus 4.6 3dfa8b3ce8 fix: promote variant spellings in word dictionary ranking
Words like "halfsister" were buried at position 99,860 (requiring Full
vocabulary) despite "half-sister" being ranked at 3,451. Strip hyphens
and spaces to normalize, then group variants together so compound forms
inherit the rank of their hyphenated/spaced counterpart.

Promotes 3,701 words. halfsister: 99,860 -> 4,020 (now in Standard).

Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-23 21:54:58 -06:00
Alex-GilbertandClaude Opus 4.6 e6b38ebc6a feat: add word frequency filtering with vocabulary size control
Sort word lists by Google Books frequency (most common first).
Add vocabulary dropdown (Common/Standard/Extended/Full) to word
mode UI that controls how many words are searched.

Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-23 14:10:08 -06:00
Alex-GilbertandClaude Opus 4.6 9d8a3c9a1f feat: add word list generation script (nouns, adjectives, plurals from WordNet)
Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-23 09:51:15 -06:00
Alex-GilbertandClaude Opus 4.6 72df8d3579 feat: extract associate first/last name lists from registry
Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-18 11:55:33 -06:00
Alex-Gilbert a5505c1fbb attempting to make redaction guessing better 2026-02-18 10:35:42 -06:00
Alex-GilbertandClaude Opus 4.6 0d88d6c0e7 feat: add build script to fetch and merge Epstein-related email addresses
Fetches from 3 data sources (HuggingFace parquet, rhowardstone entities,
Epstein Exposed API) plus 8 court-verified addresses. Merges, deduplicates,
validates, and writes 644 unique emails to unredact/data/emails.txt.

Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-16 13:29:18 -06:00
Alex-GilbertandClaude Opus 4.6 2268e8549e feat: add separate V (victim) badge independent of associate tiers
Victim names are stored as a flat set of lowercased name variants
with no person details — the V badge appears but is not clickable,
protecting victim identities. A result can show both T1/T2/T3 and V
badges simultaneously.

- 15 known victims from public court records → 43 name variants
- Enabler/perpetrator categories promoted to T2
- Purple V badge with pointer-events disabled
- Victim matching uses prefix/suffix combination like associate matching

Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-16 10:31:38 -06:00
Alex-GilbertandClaude Opus 4.6 eb42b0e179 fix: correct scoring inversion, filter garbage names, add Makefile target
- Fix priority scoring: (4 - tier) * weight so T1 ranks above T3
- Filter upstream redaction placeholders like "(b) (6)" from data
- Add build-associates Makefile target
- Rebuild associates.json (1586 persons, 28 garbage entries removed)

Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-16 09:17:48 -06:00
Alex-GilbertandClaude Opus 4.6 b5d2670fa3 feat: generate associates.json with tiered Epstein network data
Fix source_name/target_name field mapping in tier computation, use
compact JSON output (696KB). Stats: 7147 lookups, 1614 persons
(204 tier-1, 38 tier-2, 1372 tier-3).

Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-16 07:49:17 -06:00
Alex-GilbertandClaude Opus 4.6 8011cf4da2 feat: add associates data processor with nickname generation and tiering
Co-Authored-By: Claude Opus 4.6 <[email protected]>
2026-02-16 07:46:56 -06:00