Elamur Groupantsen-tag-training
The groupant beside en/src/tag.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 | approximant |
| ast | partial | ../syntax.json |
| composition | partial | composition |
| descriptant | present | descriptant |
| invariant | partial | invariant |
| tests | executed bounded checks | 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 |
interpolation_boundary | interpolation_boundary | partial source derived model | interpolation_boundary |
suffix_backoff | suffix_backoff | partial source derived model | suffix_backoff |
emission_precedence | emission_precedence | partial source derived model | emission_precedence |
viterbi_ties | viterbi_ties | partial source derived model | viterbi_ties |
Model text
descriptant
en-tag-training
Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Epoch 2026-10-31: source-scoped partial model; parent tests provide finite shared evidence.
2026-11-04 epoch (en-morph): source rebound to e2175f4ca53b; byte range 2272..6558 → 2272..6558. The scope text is unchanged (moved only).
approximant
Scoped forward relation
Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. No executable percentage approximant attached.
invariant
Scoped assumptions and costs
Empty tagging empty; empty training has NaNs/no support and nonempty inference may panic. Deserialize and Soft multiplier have no validation. Train O(T+NT^3+suffix work); decode O(nK^3) time/O(nK^2), NT fixed; unchecked counters.
composition
Parent composition
Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Parent combines this fragment with the other named source fragments and helpers outside their ranges. No coverage claim for unscoped code.
models
behavior_complement
Within this fragment's byte scope: Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Parent relation constraining compatible states: Empty tagging empty; empty training has NaNs/no support and nonempty inference may panic. Deserialize and Soft multiplier have no validation.
reverse_reconstruction
Within this fragment's byte scope: Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Parent relation constraining compatible states: A sequence has all words/parameters for which its score maximizes under strict-greater first-candidate ties; aggregated training counts lose original sentence ordering.
interpolation_boundary
Within this fragment's byte scope: Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Parent relation constraining compatible states: Deleted interpolation chooses q3 on ties then q2; no counts makes lambda division0/0 and NaN transition cells.
suffix_backoff
Within this fragment's byte scope: Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Parent relation constraining compatible states: Suffix smoothing (q+theta*p)/(1+theta) stops at first missing suffix, cap flag distinguishes uppercase first scalar.
emission_precedence
Within this fragment's byte scope: Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Parent relation constraining compatible states: Known exact/lowercase lexical entries return before dictionary; hints only modify unknown probabilities; hard cannot add unseen classes.
viterbi_ties
Within this fragment's byte scope: Tagged sentences accumulate uni/bi/tri counts including BOS. Deleted interpolation estimates lambda with trigram-first ties; transitions are log mixtures. Rare word suffix counts and theta construct unknown emissions. Empty input has no finite probability guarantee. Parent relation constraining compatible states: Viterbi sums transition+emission and final BOS; strict > retains first finite maximizing predecessor; candidate emptiness is not validated.
tests
Bounded parent evidence
See ../tests.md. Actual check: tag::wave27_tests::tiny_known_emissions_and_empty_boundary; shared source evidence, not independent exhaustive fragment testing.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/en/src/tag.rs.groupant/training.groupant/