Repository navigation
World stage b3b: numeric fluents in the corpus, RAISES and LOWERS, conditional causal statements, tense on auxiliaries - #29
Merged
Conversation
…nditional causal statements, and the tense affix on auxiliaries
The last stage of phase (b) of the world-and-language refactor (docs/specs/WORLD_AND_LANGUAGE.md, "b3. The corpus", second half), with Jon's rulings 2 and 3 on stage b3a as its first two items.
- Jon's four rulings on stage b3a are recorded at the end of its proposal file, and WM.E207 to WM.E223 are Decided. Comparison partners default to the siblings (documents.comparison_partners: siblings | thematic | both); a comparative test item draws its word, "bigger" or "smaller", half and half from the tests stream, once per pair.
- Words for numeric fluents: two absolute adjectives per fluent (concept type numeric, lexicon.named_proportion.numeric), judged against the ends of the range at numeric_adjectives.cut_high (0.75) and cut_low (0.25); a numeric pole is a comparison with RANGE at a time point and records its cut in the predicate; the adjectives take the comparative forms of stage b3a.
- Numeric states join the initial-state sentences, numeric changes the result sentences (INCREASE and DECREASE, stated when the event's own effect moved the value and the step's net change has its direction; never by drift alone), and numeric comparatives at the event's time point the comparative descriptions. Truth and the oracle judge on the replayed values.
- RAISES and LOWERS: effect statements with a direction, true exactly when every event type below has a numeric effect of the sign on the role; "become" with the comparative form; numeric fluents are topics of feature documents.
- Conditional causal statements: the condition's literals on the event term and in the restrictor after the bound event (a static literal as an atom, a fluent literal under BEFORE), realized as modifiers of the subject or of the other participant and as "before" predications of relative clauses; true exactly when every leaf has the effect under that condition; a statement without its condition is never true; conditions the grammar cannot say (a fluent or a negated PROPERTY of the patient in a statement about the patient) and numeric threshold conditions make no statement and are counted. A conditional numeric effect makes a conditional RAISES or LOWERS.
- Test sets: numeric poles in the state sets with changed marks, the new set state_comparison_role, RAISES and LOWERS in the causal sets with the direction swap as the polarity swap, and the new change condition (causal_effect_condition and its law-like twin).
- With the tense as an affix, the auxiliary of a sentence about a time point carries it; the request lists the inflected function words ({function_words, affixes} inflect entries), and the pipeline makes their forms (the one pipeline change: InflectEntry.function_words, function words drawn among forms that take their affixes, FUNCWORD.<n>.AFFIX.<m>), with content words, assignment, audio, and embeddings unchanged.
- stats.yaml counts the numeric states, changes, comparatives, RAISES, LOWERS, and conditional sentences, and the conditional effects that make no statement, by reason. The oracle of the tests judges every new form independently.
Regenerated runs: both worlds unchanged (rule sets 912f1d371460 and f47334502194); the default corpus has 100,906 sentences, 482,455 tokens, 221 lexemes, 46 sets with 16,433 pairs (99,772, 479,854, 217, 43 sets with 15,849 pairs in stage b3a); the snapshot moved 203 figures (the sentences by level, the causal, comparisons, and states blocks, the lexicon, the test sets, and the co-occurrence: the category documents' thematic correlation 0.50 to 0.11 under the sibling default, 0.17 before stage b3a). Run times: world 6.3 s (6.1 s), corpus 141 s (132.6 s), default word forms 30 min 4 s (31 min 33 s); no step grew by more than half.
The full check list passes: 2,073 tests, none skipped (fast tier 1 min 47 s, slow tier 7 min 53 s, both together 6 min 34 s). Engineering choices WM.E224 to WM.E239 (Working design) in docs/DECISIONS.md and docs/proposals/2026-10-15-world-stage-b3b-decisions.md, with five open questions for Jon and notes for phase (c).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stage b3b of the world-and-language refactor; completes phase (b). See docs/proposals/2026-10-15-world-stage-b3b-decisions.md.