Lemmas where two tagged dictionaries assert disjoint specific genders ({m, f, n}). These are review candidates, not verdicts — follow the source links to judge each case.
Deep comparison uses validated feature adapters only. Broad coverage/overlap covers eligible local Sanskrit/BHS headwords; missing deep markup is not counted as zero evidence.
Trust Block
- Evidence:
src/data/dicts/pos-disagreement.json,src/data/dicts/alignment-confidence.json, and dictionary source links. - Limitations: conflicts are machine-derived review candidates; they do not prove that one dictionary is wrong.
- Validation: generated by
npm run build-dict-comparison; checked bynpm run validate-dict-comparison,npm run validate-review-reports, andnpm run build. - Owner repo:
csl-atlas. - Next use: inspect highlighted rows, then open exact dictionary source records before citing the pattern.
Gender is derived from validated adapters:
<lex> tags, dictionary-specific prose markers, and parsed kosha synonym suffixes where tested. A conflict means the dictionaries disagree on a specific gender; adjective/indeclinable tags never trigger one. VCP under-marks feminine/neuter at the anchor position, so some VCP f/n genders are simply absent (a missed conflict, never a false one).
How confidently lemmas align across dictionaries.
Because all seven dictionaries store SLP1 headwords, cross-dictionary alignment is almost always byte-identical — alignment uncertainty is low. The harder, deferred cases are homonym splitting and sense alignment (see the comparison plan).
Generated by npm run build-dict-comparison. CC-BY-SA-4.0.