Skip to content

docs(taxonomy): #498 PR-10 Equivoque / lexical-polysemy cluster (5 leaves) #717

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jsboige merged 1 commit into from
Jul 6, 2026
Merged

docs(taxonomy): #498 PR-10 Equivoque / lexical-polysemy cluster (5 leaves) #717

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
294 changes: 294 additions & 0 deletions docs/taxonomy/498-aif-equivoque-polysemie-cluster.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,294 @@
# 2026-07-06 — #498 AIF chantier, PR-10 : cluster « Equivoque / Polysémie » (lexical multi-meaning)

**Scope**: tenth cluster of the #498 chantier — first half of the **Equivoque** sub-sub (Ambiguïté
subfamily, Abus de langage / Misleading language family). Models the **lexical polysemy mechanism**
(pk 855 anchor + 860/862/863/864 leaves, using 856 as mapped template). **Proposition only — GATED,
0 write to prod CSV.** Triggered by ai-01 dispatch `5ymo70` (primary: Equivoque sub-sub, split by
mechanism). **In-sub-sub anchor shape — independent of the I/RA/CA serialization decision.**

**Repo reference**: master `bdba45d8`. Issue: #498. Predecessors: PR-1 #699, PR-2 #701, PR-3 #703,
PR-4 #705, PR-6 #708, PR-7 #711 (Amphibologie), PR-8 #713, PR-9 #714 (**Ambiguïté subfamily complete
3/3**).

> **⚠ STATUS = PROPOSITION ONLY.** No prod CSV write, no DB write, no OWL regen. jsboige ratifies →
> worker applies the CSV edits in a follow-up PR (gated).

---

## TL;DR

The **Equivoque** sub-sub (`Soussousfamille=Equivoque`) has **2 mapped rows** — pk 856 (d4 "Expression
vague", MAPPED `VagueVerbalClassification_Inference`, direct-conflict, `skos:broadMatch`) and pk 858
(d5 "Pente glissante sémantique", MAPPED `VerbalSlipperySlope_Inference` exception-ref, `skos:closeMatch`)
— and **18 unmapped leaves** (pk 855 anchor d3 + 857, 859-875). The sub-sub is the **largest in the
chantier** (21 rows) and semantically spans **three distinct mechanisms**, so it is split by mechanism
(this PR = lexical polysemy; PR-11 = continuum/sorite + misc; PR-12 = reification/magic-of-words).

**PR-10 scopes to the lexical-polysemy mechanism**: anchor pk 855 + template pk 856 (recap) + 4 leaves
(860 glissement polysémique, 862 homonymie, 863 hétérosémie, 864 jingle-jangle). All 5 modeled rows
honestly reuse the template 856's `VagueVerbalClassification_Inference` — the lexical-polysemy defect
*is* a vague verbal classification (the term's extension is not fixed, the arguer exploits one reading).
The only variation is the `MappingType`, encoding how tightly each leaf specializes the polysemy
mechanism.

**5 rows proposed, 5 fully-modeled, 0 FAIL-LOUD** — the third consecutive clean cluster (after PR-8,
PR-9). This PR opens the Equivoque sub-sub; the continuum (PR-11) and reification (PR-12) mechanisms
are deferred.

---

## 1. Cluster state (code=truth)

From `Cards/Fallacies/Argumentum Fallacies - Taxonomy.csv`, family **Abus de langage** (Misleading
language), subfamily **Ambiguïté**, sub-sub **Equivoque**:

| pk | text_fr | depth | DirectRef | ExceptionRef | MappingType | State |
|----|---------|-------|-----------|--------------|-------------|-------|
| 855 | Équivoque | d3 | — | — | — | unmapped (d3 anchor, **PR-10**) |
| **856** | **Expression vague** (mapped template) | d4 | `VagueVerbalClassification_Inference` | — | `skos:broadMatch` | **✅ MAPPED (PR-10 template)** |
| 860 | Glissement lexical polysémique | d4 | — | — | — | unmapped (**PR-10**) |
| 862 | Homonymie | d5 | — | — | — | unmapped (**PR-10**) |
| 863 | Hétérosémie | d6 | — | — | — | unmapped (**PR-10**) |
| 864 | Sophisme jingle-jangle | d5 | — | — | — | unmapped (**PR-10**) |

The mapped template (856) is the **paradigm**: an equivocal term = a `VagueVerbalClassification_Inference`
(a verbal classification where the term's extension is not fixed, so the arguer can exploit any reading).
The lexical-polysemy leaves (855, 860, 862, 863, 864) are all specializations of this — they vary by
*how* the multi-meaning arises (drift / unrelated-homonymy / context-dependence / bidirectional
confusion). The other mechanisms (continuum 858/857/859, reification 867-875) are deferred to PR-11/12.

---

## 2. The template model (856, the paradigm)

**Scheme**: `VagueVerbalClassification_Inference` (AIF native, confirmed 2x in existing usage — used as
PR-4 anchor too). A verbal classification (grouping/labeling items via a term) where the term's
extension is not fixed, allowing the arguer to shift reading mid-discussion.

**desc_fr** (856): "Vous utilisez des termes si vagues qu'ils ne permettent pas de comprendre
clairement ce que vous voulez dire."

**Pattern**: direct-conflict (the scheme is in DirectRef, no ExceptionRef — the equivocation *is* the
vague classification). This is the same shape as PR-4's anchor 800 (`VagueVerbalClassification_Inference`
direct-conflict) — the Equivoque sub-sub's lexical-polysemy mechanism is the term-level cousin of the
Inexact-definition subfamily's vagueness mechanism. The two sub-subs honestly share the scheme; the
distinction is *family* (ambiguity vs definition), not *scheme*.

---

## 3. Proposed AIF structure for the leaves

For each leaf: reuse the template 856's scheme `VagueVerbalClassification_Inference`, vary the
`MappingType` by the leaf's specialization. Vocabulary restricted to AIF-native tokens.

### pk 855 — Équivoque (Equivocation, d3 anchor) ✅
- **desc_fr**: "Vous exploitez les différents sens d'un même mot pour rendre votre propos ambigu."
- **Mechanism**: the *core* of the template — a term with multiple senses, exploited to keep the
argument ambiguous. The anchor names the whole sub-sub's mechanism in general terms.
- **Proposal**: `DirectRef=VagueVerbalClassification_Inference`, `ExceptionRef=` *(direct-conflict
pattern, mirroring 856)*, `MappingType=skos:closeMatch` (the anchor *is* the abstract case of the
template 856 — direct sibling, framing the whole sub-sub).

### pk 860 — Glissement lexical polysémique (Polysemic drift) ✅
- **desc_fr**: "Vous employez un mot polysémique ou dont le sens a évolué, ce qui peut prêter à
confusion."
- **Mechanism**: polysemy by **active drift** — the arguer slides between the senses of a polysemic
term mid-argument. The *active* form of equivocation (the arguer exploits the drift), as opposed to
862 (passive existence of homonymy).
- **Proposal**: `DirectRef=VagueVerbalClassification_Inference`, `ExceptionRef=`,
`MappingType=skos:closeMatch` (a direct variant: polysemy exploited *via active drift mid-argument*,
the canonical equivocation case — closest sibling of 855).

### pk 862 — Homonymie (Homonymy) ✅
- **desc_fr**: "Vous construisez votre argument en utilisant un terme à plusieurs sens, ce qui sème la
confusion."
- **Mechanism**: polysemy by **unrelated-homonymy** — the term has multiple unrelated senses (e.g.
"voler" = steal / fly), and the arguer trades on the coincidence. A *broader* phenomenon than 860's
active drift: any homonymic term, not just actively-exploited drift.
- **Proposal**: `DirectRef=VagueVerbalClassification_Inference`, `ExceptionRef=`,
`MappingType=skos:broadMatch` (a broader case: any homonymic term admitted, not just
actively-drifted polysemy — the broadest lexical-polysemy leaf).

### pk 863 — Hétérosémie (Heterosemy) ✅
- **desc_fr**: "Vous basez votre argument sur un mot qui peut signifier plusieurs choses selon le
contexte."
- **Mechanism**: polysemy by **context-dependence** — the term's meaning shifts with context, and the
arguer suppresses the context shift. A *narrower* specialization than 862: the meanings are related
(heterosemy, not homonymy), and the defect is the *suppressed context*, not the unrelated senses.
- **Proposal**: `DirectRef=VagueVerbalClassification_Inference`, `ExceptionRef=`,
`MappingType=skos:narrowMatch` (a narrower case: polysemy *specifically via suppressed context*, the
heterosemy specialization — narrower than 862's broad homonymy).

### pk 864 — Sophisme jingle-jangle (Jingle-jangle fallacy) ✅
- **⚠ Bidirectional confusion**: unlike 860/862/863 (one term, multiple meanings), the jingle-jangle
fallacy goes *both ways* — treating different terms as the same idea (jingle) OR the same term as
different ideas (jang). The desc says it explicitly: "termes différents pour désigner la même idée,
ou le même terme pour désigner des idées différentes". This is the **mirror** of the others — it
confuses the term↔idea mapping in both directions.
- **Mechanism**: the `VagueVerbalClassification_Inference` still applies (the classification is vague
because the term↔idea mapping is inconsistent), but the bidirectional confusion is the
*characteristic* specialization.
- **Proposal**: `DirectRef=VagueVerbalClassification_Inference`, `ExceptionRef=`,
`MappingType=skos:closeMatch` (a direct variant: equivocation *via bidirectional term↔idea
confusion*, a direct mirror-sibling of the anchor 855's unidirectional case).

---

## 4. Cluster summary (proposed)

| pk | text_fr | Proposed ExceptionRef | Proposed DirectRef | MappingType | Honest? |
|----|---------|----------------------|-------------------|-------------|---------|
| 855 | Équivoque | *(absent)* | `VagueVerbalClassification_Inference` | `skos:closeMatch` | ✅ (anchor) |
| 860 | Glissement lexical polysémique | *(absent)* | `VagueVerbalClassification_Inference` | `skos:closeMatch` | ✅ |
| 862 | Homonymie | *(absent)* | `VagueVerbalClassification_Inference` | `skos:broadMatch` | ✅ |
| 863 | Hétérosémie | *(absent)* | `VagueVerbalClassification_Inference` | `skos:narrowMatch` | ✅ |
| 864 | Sophisme jingle-jangle | *(absent)* | `VagueVerbalClassification_Inference` | `skos:closeMatch` | ✅ |

**5 rows proposed, 5 fully-modeled, 0 FAIL-LOUD.** The third consecutive clean cluster (after PR-8,
PR-9). All leaves reuse the template 856's `VagueVerbalClassification_Inference` (direct-conflict
pattern, mirroring 856/PR-4 800). The only variation is the `MappingType`, encoding the polysemy
specialization: 862 broad (any homonymy), 863 narrow (suppressed context heterosemy), 855/860/864
close (direct variants: abstract / active-drift / bidirectional-mirror).

### Polysemy-mechanism taxonomy (§5 lesson)
Within the single `VagueVerbalClassification_Inference` scheme, the lexical-polysemy leaves specialize
along **three orthogonal axes**:
- **Directionality**: unidirectional (855/860/862/863 — one term, many meanings) vs bidirectional
(864 jingle-jangle — term↔idea confusion both ways).
- **Relation between meanings**: unrelated homonymy (862 — broad) vs related heterosemy (863 —
narrow).
- **Activeness**: passive existence of multi-meaning (862/863) vs active drift mid-argument (860).

The MappingType encodes the *breadth* (broad/narrow), while the desc captures the *axis*. This gives
the consumer a graded notion of polysemy, not a binary one — same pattern as PR-4's vagueness-mechanism
taxonomy.

---

## 5. Method notes (additions for the chantier)

- **Mechanism-based sub-sub split (large sub-sub)**: the Equivoque sub-sub (21 rows) is the largest in
the chantier. Splitting by AIF-relevant mechanism (lexical polysemy / continuum-sorite / reification)
produces 3 coherent PRs rather than one incoherent 18-leaf PR. Each mechanism maps to a distinct
AIF-scheme family (`VagueVerbalClassification_Inference` / `VerbalSlipperySlope_Inference` /
reification-needs-new-CQ), so the split is semantically clean. This generalizes PR-8/9's
Ambiguïté-narrative mechanism split to an even larger sub-sub.
- **Cross-subfamily scheme reuse (856 → PR-4)**: the template 856 `VagueVerbalClassification_Inference`
is the same scheme as PR-4's anchor 800 (Vague definition). The Equivoque sub-sub (Ambiguïté
subfamily) and the Vague-definition sub-sub (Inexact-definition subfamily) honestly share the scheme
— the distinction is *family* (ambiguity = multiple readings, definition = unclear meaning), not
*scheme*. This is cross-**subfamily** reuse (one level up from PR-6 831 / PR-7 854 cross-**sub-sub**
reuse), documented.
- **Bidirectional-confusion specialization (864)**: the jingle-jangle fallacy is the chantier's first
*bidirectional* equivocation (term↔idea both ways). The `VagueVerbalClassification_Inference` scheme
captures it (the mapping is inconsistent), but the bidirectionality is the characteristic axis,
recorded in the desc not the token. Same discipline as PR-4 §5 "intrinsic contestability is
structural, captured in desc not token".

---

## 6. DoD status

| DoD item (#498) | Status |
|-----------------|--------|
| Effective cluster-level coverage documented (not "100% leaves") | ✅ §4 (5 lexical-polysemy rows fully-modeled; Equivoque sub-sub split documented, ~13 leaves deferred to PR-11/12) |
| Per unmapped leaf: legitimate scheme + exception/CQ (not Latin alone) | ✅ §3 (5 leaves, scheme + mechanism each, 0 FAIL-LOUD) |
| `AIF_skosMappingType` coherent | ✅ §4 (close/broad/narrow encode the polysemy-breadth axis) |
| Fail-loud when no honest scheme fits | N/A this cluster (all leaves reuse the template's native scheme) |
| OWL regen reflects new structures | ⏳ Deferred — jsboige ratifies, then CSV edits, then OWL regen (#133) |

### Coverage accounting (cumulative, chantier-wide — dual criterion)
| PR | Cluster | Leaves | Fully-modeled (strict) | DirectRef-loose |
|----|---------|--------|------------------------|-----------------|
| baseline | — | — | 70 | 70 |
| PR-1 #699 | False analogy | 4 | 3 (840 CA-missing)¹ | 74 (+4) |
| PR-2 #701 | Faulty comparison | 5 | 2 (834/835/837 RA-missing) | 79 (+5) |
| PR-3 #703 | Association fallacy | 2 | 2 | 81 (+2) |
| PR-4 #705 | Vague definition | 3 | 3 | 84 (+3) |
| PR-6 #708 | Inconsistent definition | 7 | 5 (829 CA-missing, 832 RA-missing) | 89 (+5) |
| PR-7 #711 | Amphibologie | 8 | 5 (847/848/853 CA-missing) | 94 (+5) |
| PR-8 #713 | Narrative ambiguity / insinuation | 4 | 4 | 98 (+4) |
| PR-9 #714 | Narrative ambiguity / deception | 6 | 6 | 104 (+6) |
| **PR-10 (this)** | **Equivoque / polysemy** | **5** | **5** | **109 (+5)** |
| **total (strict)** | **4 subfamilies touched** | **44** | **105** | **109** |
| **total (inclusive²)** | | **44** | **106** | **109** |

¹ PR-1 originally reported "+4 mapped" counting pk 840 (RA-exists/CA-missing) as mapped. Under the
strict criterion (CA must exist), PR-1 = +3. Flagged by po-2024 audit #709.
² Inclusive = count RA-exists/CA-missing rows (840) as fully-modeled (PR-1's original convention).
The 1-row discrepancy (105 strict vs 106 inclusive) is the accounting decision pending jsboige.

**This PR's +5 is criterion-independent** — all 5 leaves have both a scheme and a conflict.

**Ambiguïté subfamily: Equivoque sub-sub opened (5/~18 leaves done)**. Remaining Equivoque mechanisms:
continuum/sorite (858 template + 857/859, ~3 leaves → PR-11), reification/magic-of-words (867-875,
9 leaves → PR-12), plus misc (861 mondegreen, 865 changement sémantique, 866 métonymie → PR-11).

---

## 7. Mapping to jsboige's enriched DoD (I-node / RA-node / CA-node decomposition)

### pk 855 — Équivoque → **undercut** (degenerate)
- **RA-node**: `VagueVerbalClassification_Inference` — classifying via a term whose extension is not
fixed.
- **CA-node**: the scheme itself names the defect (direct-conflict pattern) — the classification is
vague because the term has multiple senses. Degenerate **undercut** (CA coincides with RA, as in
PR-4 800).

### pk 860 — Glissement lexical polysémique → **undercut** (degenerate)
- **RA-node**: `VagueVerbalClassification_Inference`.
- **I-nodes**: premise P1 (at point A) "term T means M1"; premise P2 (at point B) "term T means M2";
conclusion C "items classified via T at B are grouped as at A".
- **CA-node**: degenerate **undercut** — the inference (that T's classification transfers from A to B)
is defective because P1 and P2 use different senses of T (active drift). The conclusion may hold;
the *inference* is defective.

### pk 862 — Homonymie → **undercut** (degenerate)
- **RA-node**: `VagueVerbalClassification_Inference`.
- **CA-node**: degenerate **undercut** — T has unrelated senses, so the classification has no fixed
extension.

### pk 863 — Hétérosémie → **undercut** (degenerate)
- **RA-node**: `VagueVerbalClassification_Inference`.
- **CA-node**: degenerate **undercut** — T's meaning shifts with context, suppressed by the arguer.

### pk 864 — Sophisme jingle-jangle → **undercut** (degenerate)
- **RA-node**: `VagueVerbalClassification_Inference`.
- **CA-node**: degenerate **undercut** — the term↔idea mapping is inconsistent (bidirectional).

### Cluster attack-type distribution
| pk | RA-node | CA-node | Attack type |
|----|---------|---------|-------------|
| 855 | `VagueVerbalClassification_Inference` | *(scheme as direct conflict)* | undercut (degenerate) |
| 860 | `VagueVerbalClassification_Inference` | *(scheme as direct conflict)* | undercut (degenerate) |
| 862 | `VagueVerbalClassification_Inference` | *(scheme as direct conflict)* | undercut (degenerate) |
| 863 | `VagueVerbalClassification_Inference` | *(scheme as direct conflict)* | undercut (degenerate) |
| 864 | `VagueVerbalClassification_Inference` | *(scheme as direct conflict)* | undercut (degenerate) |

**All 5 leaves are degenerate undercuts** — the direct-conflict pattern (scheme as its own defect), as
in PR-4's anchor 800 and PR-9's 885. The lexical-polysemy fallacies *are* defective
`VagueVerbalClassification_Inference`s; the inference (term ⟹ classification) is defective because the
term's extension is not fixed. No undermine, no rebut.

### Representation note
As in PR-1 to PR-9, the I-node/RA-node/CA-node decomposition is **recorded here but not serialized**
in the CSV — pending jsboige's schema decision (§4 of the coverage-status note #707). This PR is
**in-sub-sub template and independent of that decision**: the modeling (schemes + conflicts) holds
regardless of whether I/RA/CA is serialized in new columns.

---

## Gate boundaries (HARD — proposition only)

- ❌ No prod CSV write, no DB write, no OWL regen, no `aif:` assertion emitted.
- ❌ No #674/#596/#666/#686 merge, no régén launch.
- ✅ Proposition derived code=truth from the taxonomy CSV + AIF vocabulary confirmed by existing
usage (the only token proposed — `VagueVerbalClassification_Inference` — is reused from the template
856, no fabrication).
- ✅ Mechanism-based sub-sub split documented; cross-subfamily scheme reuse (856 → PR-4) noted;
bidirectional-confusion specialization (864) documented; accounting transparency (dual criterion,
§6) provided re: po-2024 audit #709.

Relates: dispatch `5ymo70` (primary), #498 (chantier), PR-1 #699, PR-2 #701, PR-3 #703, PR-4 #705,
PR-6 #708, PR-7 #711, PR-8 #713, PR-9 #714, coverage-status #707, verification-audit #709,
family-gap #712, #133/#130 (existing OWL), #499 (inverse), #677 (0 fabrication), #192, #458.
Loading