Skip to content

todo.md

Latest commit

 

History

History
27 lines (21 loc) · 1.46 KB

todo.md

File metadata and controls

27 lines (21 loc) · 1.46 KB
Raw

Active plan — Morpheus × LXX

Completed rounds 1–6 live in completed.md: stemlib coverage to 99.81%, the full-corpus annotation pipeline, the repo-wide scrub, the docs refresh, and the Round-6 convention layer (dictionary forms + base lemmas) that took the LXX annotation past the reference on all three metrics. Working files for this work live in tmp/lxx/ (git-ignored): compare.py, eval_morph.py, report.txt, TSVs, and morph-lxx/ (the annotated corpus). Round 6 added analysis tools (diff_variants.py, gen_tiebreak.py, table_audit.py, probes), per-variant outputs (morph-lxx-base2/, morph-lxx-dict/, morph-lxx-oldnat/) and eval snapshots (eval_r6*.txt).

Next round — candidates (not yet approved; plan before implementing)

No active tasks. Candidate directions from the Round-6 review:

  1. Port the convention layer to NT — lemma_frequencies.json already carries joint lemma×POS counts, so the ranking floor applies there for free; a build_morph_gnt.py with dictionary-form + base-lemma modes (tables derived from an NT reference corpus) would be the natural Round 7. Needs a local checkout of that reference corpus.
  2. C# port parity — lessons.md notes placeholder stubs (chckindecl.cs returns 0; CheckGenWordsFunc lacks comptab/accent logic); worth fixing if any deployment path uses the C# runner.
  3. Small LXX polish (ἕως/ἠώς homograph, ~277 tokens) — diminishing returns; optional.