Skip to content

Add Amharic — third Semitic engine, Ge'ez script (Beta) - #153

Merged
v4nn4 merged 1 commit into
mainfrom
rf/amharic-engine
Aug 23, 2026
Merged

v4nn4 merged 1 commit into
mainfrom
rf/amharic-engine

Conversation

@v4nn4

@v4nn4 v4nn4 commented Aug 23, 2026

Copy link
Copy Markdown
Collaborator

Third engine in the Semitic push, and the hardest script — the Ge'ez abugida.

Approach

Amharic is cited by the 3sg-masc perfective (ሄደ). The engine resolves each lemma to six mined principal parts (perfective, imperfective, perfect, present, imperfective-nonfinite, 2sg-masc imperative) and derives the paradigm by regular subject-agreement affixation per TAM. A Ge'ez syllable helper decomposes each Ethiopic codepoint into (consonant-row, vowel-order) so stem vowels can be revocalised and agreement syllables appended. A 408-cell mined-override layer patches the irregular residue (quadriliteral reduplications, geminate/biradical roots, type-B ablaut, causative/glottal prefix merges) and seeds the 17 part-less lemmas.

Verification

Beta — single oracle (UniMorph; kaikki's Amharic doesn't expand). Golden gate: forms 20247/20247 (100.00%), lemma coverage 670/670 (100.00%). 97.98% rule-derived, 2.02% overridden.

Registration

Lang::Amh, from_code am/amh/amharic, Conjugation::Amh + build dispatch, reverse index, a 5-group serde Table, PyO3 + wasm, Cargo includes, correctness.py, CI single-oracle gate. All green: 343 lib tests, clippy -D warnings, fmt, --features python/--features wasm.

Stacked on rf/hebrew-engine (#152) — merge that first. Next: web wiring for both, then Syriac / Egyptian / Maltese.

🤖 Generated with Claude Code

Base automatically changed from rf/hebrew-engine to main August 23, 2026 22:43
Amharic (Ethiosemitic) is written in the Ge'ez abugida and cited by the 3sg-masc
perfective (ሄደ). The engine resolves each lemma to six mined principal parts
(perfective, imperfective, perfect, present, imperfective-nonfinite and the
2sg-masc imperative; data/amh/parts.tsv) and derives the paradigm by regular
subject-agreement affixation per TAM, with a Ge'ez syllable helper that
decomposes each Ethiopic codepoint into (consonant-row, vowel-order) so stem
vowels can be revocalised and agreement syllables appended. Weak/irregular
residue (quadriliteral reduplications, geminate/biradical roots, type-B ablaut,
causative/glottal prefix merges) is patched by a mined-override layer
(data/amh/overrides.tsv, 408 cells — 159 rule residue + 249 seeding the 17
part-less lemmas incl. the copula ነው).

Beta tier: verified against a single oracle, UniMorph (kaikki's Amharic
extraction does not expand). Golden gate: forms 20247/20247 (100.00%), lemma
coverage 670/670 (100.00%), 97.98% rule-derived / 2.02% overridden.

Full registration: Lang::Amh, from_code am/amh/amharic, Conjugation + build
dispatch, reverse index, a 5-group serde Table (perfective/perfect/imperfective/
imperfectiveNfin/jussive), PyO3 + wasm, Cargo includes, correctness.py, and the
CI single-oracle gate. All green: 343 lib tests, clippy, fmt, python + wasm.

Stacked on the Hebrew branch (rf/hebrew-engine); merge that first.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@v4nn4
v4nn4 force-pushed the rf/amharic-engine branch from 2e8caca to a409941 Compare August 23, 2026 22:44
@v4nn4
v4nn4 enabled auto-merge (squash) August 23, 2026 22:44
@v4nn4
v4nn4 merged commit 32455d1 into main Aug 23, 2026
2 checks passed
@v4nn4
v4nn4 deleted the rf/amharic-engine branch August 23, 2026 22:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant