Elamur Groupantsmova-nautilus-senses
The groupant beside tools/nautilus/src/senses.rs in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.
Facts
Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.
Components
| component | label | where |
|---|---|---|
| approximant | partial source derived | Forward approximant |
| ast | partial | syntax.json |
| composition | partial source derived | Composition |
| cost | partial source derived | Cost |
| descriptant | present | Descriptant |
| invariant | partial source derived | Invariant |
| tests | finite evidence | tests |
Obernants
The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.
| id | type | label | where |
|---|---|---|---|
behavior_complement | behavior_complement | partial source derived model | behavior_complement |
reverse_reconstruction | reverse_reconstruction | partial source derived model | reverse_reconstruction |
Model text
model
mova-nautilus-senses
Descriptant
Senses: which of the pupils' numbered senses a layer's examples show, natively first. tokens/lemmas (en::dict analyses, comparatives only from UD, archaic verb endings), find (lemma match, never prefixes), junk (< 3 letters, scan abbreviations), boilerplate examples, uses, excerpt (~160 chars around the word, no id), parse (numbered senses), content lemmas, score (shared lemmas, quoted phrases x3), native_pick (one sense and >= 1 shared, or best >= 2 and >= next + 2), restates (Dice >= 0.5 and >= 3 shared), absent, Pick (numbers, new, same, -), Store (senses.jsonl: first numbered/new pick fixes a word), plan/WordPlan (Junk, Thin, Native, Ask; block; native_line), pick_line, records. Source: tools/nautilus/src/senses.rs.
Forward approximant
plan(word, examples, min_uses, senses, fixed, native_ok): junk(word) -> Junk; uses = examples using word by lemma, not boilerplate; |uses| < min_uses -> Thin; !native_ok or no senses -> Ask; native_pick -> Native(pick, sense text, evidence; agrees/differs with fixed) else Ask. Store.fixed = first record per word with a numbered or new pick.
Invariant
Deterministic: BTreeSet/BTreeMap, ties by sense number; a native pick shares >= 1 lemma with its sense; a restated text shares >= 3 lemmas and Dice >= 0.5 with its sense; the store is append-only.
Cost
O(tokens x analyses) per example; one store read per plan batch; appends one line per pick.
Composition
Parent: nautilus. Used by work::DictJob, improve::LayerJob, teacher::dictionary_task, replay.
behavior_complement
Never matches a word by prefix; never asks junk or thin words; never writes a layer or admits; a native pick needs the stated overlap; the store records are review-only (admitted false). Evidence status: source-derived model with finite checks, not a proof.
reverse_reconstruction
A pick line or store record recovers word, pick, sense text, layer, by and evidence; the examples' shared lemmas are named in the evidence. Unique recovery of the examples is not claimed.
Epoch 2026-11-04
First model, written with the source; prospective: on new runs the native share stays near the replay's 10% and restatements near 10% of written senses.
tests
Checks
Command: cargo test --release --manifest-path tools/nautilus/Cargo.toml --offline --locked with MOVA_OCTOPUS_TEST_MODEL=$MOVA_DATA/download/0/0.4/en.bin.
2026-11-04: words_are_matched_on_lemmas_not_prefixes (red control: a prefix match takes «aged» for «ag»), junk_and_boilerplate, senses_scores_picks_and_restatements, picks_parse_and_excerpts_cut_around_the_word, the_store_fixes_a_word_once. Red: prefix match, junk length, restatement gate (each red, restored). Finite fixtures; no teacher or model was called; not a proof.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/tools/nautilus/src/senses.rs.groupant/