SG-ENGLISH Master Graph v0.1 (Primary → Secondary → JC) — Executable Syllabus Lattice

TITLE: SG-ENGLISH Master Graph v0.1 (Primary → Secondary → JC) — Executable Syllabus Lattice
VERSION: CivOS Unified Spec v1.x
MODE: Almost-Code / LLM-first / WordPress paste-ready
OWNER: eduKateSG Fence Learning Systems
SCOPE: Singapore English (Primary + Secondary + JC) as a capability lattice
GOAL: Turn “English syllabus” into a computable system: Meaning → Sentence → Paragraph → Composition → Transfer under load
CANONICAL CLAIM:
- English failure is rarely “lack of effort”. It is coordination drift:
  (1) weak input→output pipeline,
  (2) unstable meaning under load,
  (3) vocabulary retrieval failure,
  (4) narrative fracture (NIT crossing),
  (5) late fencing (repair arrives after collapse).
- This graph attaches sensors + Fence rules + repair routing so learners reach P2/P3 reliability.
========================================================
0) GLOBAL PRIMITIVES (LOCKED)
========================================================
Phase (Reliability Under Load):
- P0 collapse: cannot produce coherent output
- P1 scaffolded: can do guided tasks; fails under variation
- P2 reliable: normal timed exam conditions stable
- P3 stable: high variation + pressure; strong voice + control
Zoom:
- Z0 learner, Z1 family/tutor, Z2 school, Z3 city, Z4 nation, Z5 civilisation, Z6 supranational
FenceOS:
- R = Ḋ/Ġ  (language decay / repair rate)  ; stable if R<1
- Θ = T_fence / T_repair                  ; must be <1
- Λ = T_enforce / T_fence                 ; preferred <1
English-HD Core Sensors:
- RR (Retrieval Reliability): % of active words usable in sentences on demand
- GR (Grammar Reliability): error rate per 100 words under timing
- CR (Coherence Reliability): paragraph logic stability
- SR (Structure Reliability): planning→execution stability (e.g., 5-Box plan)
- NIT (Narrative Irreversibility Threshold): point where story/explanation becomes irrecoverably wrong under load
- TR (Transfer Reliability): performance stability across prompts/genres
- SII (Speed Instability): output rate variance under time pressure
- AvoidSlope: d(Avoidance)/dt (avoid writing, avoid reading)
Fence Actions:
- Truncation: reduce cognitive load; fix meaning layer before style
- Stitching: targeted repair loop; rebuild buffer; regain stable trajectory (APRC)
========================================================
1) GRAPH FORMALISM (HOW TO READ THIS)
========================================================
English is a pipeline lattice:
Input Lattice → Meaning Lock → Sentence Engine → Paragraph Engine → Composition Engine → Exam Transfer
Node := capability unit
Edge := prerequisite dependency + coupling to other subskills
NodeRecord := {
  ID, LevelBand, Domain, Prereq[], Outputs[],
  LoadProfile, PhaseTargetByStage,
  TransferTests[], Sensors[], FailureModes[], RepairLoops[],
  FenceRules
}
ID Grammar (LOCK):
Place×Lane×Zoom×Role×Type×ID
Place: SG
Lane: ENG
Zoom: Z0
Role: Learner / Parent / Tutor / Teacher / Operator
Type: Node / Edge / SensorPack / Protocol / Test / Panel
========================================================
2) MASTER DOMAINS (TOP-LEVEL CLUSTERS)
========================================================
Primary Domains:
- READ: Reading & comprehension base
- VOC: Vocabulary lattice (active retrieval)
- GRAM: Grammar + sentence correctness
- SENT: Sentence craft (variety + connectors)
- PARA: Paragraph craft (topic sentence + support)
- COMP: Composition engine (planning, story, editing)
- ORAL: Oral & listening (optional but coupled)
- SP: Spelling & punctuation
Secondary Domains:
- COMP-R: Reading comprehension (literal→inferential)
- LANG: Language use (editing, synthesis, summary, situational writing)
- WRIT: Continuous writing (narrative/descriptive/argumentative basics)
- VOC2: Academic vocabulary + collocations
- STYLE: voice, tone, audience control
JC Domains:
- AQ: argumentation/analysis (essay logic)
- SYN: synthesis/summary (compression under constraints)
- EVID: evidence reasoning + critique
- DISC: discourse control (structure, thesis, rebuttal)
- EXAM: high-load timed performance
========================================================
3) PRIMARY (P1–P6) — NODE SET v0.1
========================================================
----------------------------------------
3.1 INPUT ENGINE (READING → COMPREHENSION)
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×READ-FLUENCY
Prereq: []
Outputs: READ-COMP-LIT
LoadProfile: medium (speed + accuracy)
PhaseTarget: {P2 by P3, P3 by P5 top band}
TransferTests: {read aloud; timed passage}
Sensors: {SII-reading, AvoidSlope}
FailureModes: {SkipReading, Guessing}
RepairLoops: {daily short reads + retell}
FenceRules:
- If AvoidSlope rising → Truncate length + increase frequency (micro-dose reading)
Node: SG×ENG×Z0×Learner×Node×READ-COMP-LIT
Prereq: [READ-FLUENCY]
Outputs: READ-COMP-INF
LoadProfile: medium
PhaseTarget: {P2 by P4}
TransferTests: {who/what/when/where; locate evidence}
Sensors: {TR-comp}
FailureModes: {EvidenceBlindness}
RepairLoops: {underline→quote→answer protocol}
FenceRules: enforce evidence step (Λ control)
Node: SG×ENG×Z0×Learner×Node×READ-COMP-INF
Prereq: [READ-COMP-LIT]
Outputs: LANG-USE-EDIT, COMP-PLAN
LoadProfile: high (inference + ambiguity)
PhaseTarget: {P2 by P6, P3 top band}
TransferTests: {why/how questions; tone; intention}
Sensors: {TR-comp, NIT-lite}
FailureModes: {MakeUpAnswers, NoTextEvidence}
RepairLoops: {two-evidence rule + “because-line”}
FenceRules:
- If NoTextEvidence repeated → Truncate inference → Stitch evidence retrieval
----------------------------------------
3.2 VOCABULARY LATTICE (ACTIVE RETRIEVAL)
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×VOC-ACTIVE-100
Prereq: [READ-FLUENCY]
Outputs: SENT-CORE, COMP-DESCRIPT
LoadProfile: high (retrieval under time)
PhaseTarget: {P2 by P4, P3 by P6 top band}
TransferTests: {use 10 words in 10 sentences; cloze; short paragraph}
Sensors: {RR, TR, SII}
FailureModes: {LabelOnlyTrap, PassiveOnly}
RepairLoops: {Fence Method: word→phrase→sentence→paragraph}
FenceRules:
- If RR<60% → Truncate new words; Stitch retrieval drills before expanding list
Node: SG×ENG×Z0×Learner×Node×VOC-COLLOCATION
Prereq: [VOC-ACTIVE-100]
Outputs: SENT-VARIETY, STYLE-VOICE
LoadProfile: high (naturalness)
PhaseTarget: {P2 by P6}
TransferTests: {correct collocations; fix awkward lines}
Sensors: {RR, GR}
FailureModes: {DirectTranslation, AwkwardChunks}
RepairLoops: {phrase bank + substitution drills}
FenceRules: enforce phrase-first writing for weak RR
----------------------------------------
3.3 GRAMMAR + SENTENCE ENGINE
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×GRAM-CORE
Prereq: []
Outputs: SENT-CORE
LoadProfile: medium→high (under timing)
PhaseTarget: {P2 by P5}
TransferTests: {editing; error hunt; rewrite}
Sensors: {GR}
FailureModes: {TenseDrift, SVA errors, Fragment/Run-on}
RepairLoops: {micro-rules + rewrite of one paragraph daily}
FenceRules:
- If GR worsens under timing → Truncate length + Stitch rule packs
Node: SG×ENG×Z0×Learner×Node×SENT-CORE
Prereq: [GRAM-CORE, VOC-ACTIVE-100]
Outputs: SENT-VARIETY, PARA-CORE
LoadProfile: high (clarity + correctness)
PhaseTarget: {P2 by P6}
TransferTests: {combine sentences; expand/condense}
Sensors: {GR, CR}
FailureModes: {ChoppyWriting, OverlongSentences}
RepairLoops: {sentence patterns + connectors}
FenceRules: cap sentence length for P1→P2 learners
Node: SG×ENG×Z0×Learner×Node×SENT-VARIETY
Prereq: [SENT-CORE, VOC-COLLOCATION]
Outputs: PARA-CORE, COMP-VOICE
LoadProfile: high
PhaseTarget: {P2 by P6, P3 top band}
TransferTests: {use 6 connector types; vary openings}
Sensors: {CR, SR}
FailureModes: {ConnectorBlindness, RepetitiveStarts}
RepairLoops: {connector ladders + imitation practice}
FenceRules: if ConnectorBlindness → Truncate style; Stitch connectors only
----------------------------------------
3.4 PARAGRAPH ENGINE
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×PARA-CORE
Prereq: [SENT-CORE, READ-COMP-LIT]
Outputs: COMP-PLAN, LANG-USE-SHORT
LoadProfile: high (logic + support)
PhaseTarget: {P2 by P6}
TransferTests: {topic sentence + 3 supports; PEEL lite}
Sensors: {CR, NIT-lite}
FailureModes: {PointNoSupport, DriftMidParagraph}
RepairLoops: {one-paragraph daily + checklist}
FenceRules:
- If DriftMidParagraph repeats → Truncate length → Stitch paragraph skeleton
----------------------------------------
3.5 COMPOSITION ENGINE (PRIMARY)
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×COMP-PLAN-5BOX
Prereq: [PARA-CORE, READ-COMP-INF]
Outputs: COMP-DRAFT, COMP-REWRITE
LoadProfile: high (planning under time)
PhaseTarget: {P2 by P6, P3 top band}
TransferTests: {plan in 8 minutes; execute in 35}
Sensors: {SR, SII}
FailureModes: {NoPlanRush, OverPlanNoWrite}
RepairLoops: {timed planning drills + “plan→write 1 paragraph immediately”}
FenceRules:
- If SR low → enforce plan-before-write (Λ control), but truncate plan complexity
Node: SG×ENG×Z0×Learner×Node×COMP-DRAFT
Prereq: [COMP-PLAN-5BOX, VOC-ACTIVE-100]
Outputs: COMP-EDIT, COMP-VOICE
LoadProfile: extreme (timed continuous writing)
PhaseTarget: {P2 by P6}
TransferTests: {full Paper 1 simulation}
Sensors: {SII, CR, GR, NIT}
FailureModes: {MidStoryCollapse, IncoherentShift, TimeCrash}
RepairLoops: {weakest-paragraph rewrite + timed ramp}
FenceRules:
- If NIT triggered → TRUNCATE plot complexity → STITCH coherence via 3-scene spine
Node: SG×ENG×Z0×Learner×Node×COMP-EDIT
Prereq: [COMP-DRAFT, GRAM-CORE]
Outputs: COMP-REWRITE
LoadProfile: high (self-correction under time)
PhaseTarget: {P2 by P6}
TransferTests: {edit checklist in 6 minutes}
Sensors: {GR}
FailureModes: {CannotSeeOwnErrors}
RepairLoops: {read-aloud edit + error class tags}
FenceRules: must edit 1 paragraph minimum every cycle
Node: SG×ENG×Z0×Learner×Node×COMP-VOICE
Prereq: [SENT-VARIETY, VOC-COLLOCATION]
Outputs: P3 writing
LoadProfile: high (style without losing clarity)
PhaseTarget: {P3 by P6 top band}
TransferTests: {rewrite same scene in 2 tones}
Sensors: {CR}
FailureModes: {PurpleProse, StyleOverMeaning}
RepairLoops: {clarity-first rewrite}
FenceRules: if clarity drops → truncate style features first
========================================================
4) SECONDARY — NODE SET v0.1
========================================================
----------------------------------------
4.1 COMPREHENSION + LANGUAGE USE
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×COMP-SEC-INFERENCE
Prereq: [READ-COMP-INF]
Outputs: COMP-SEC-EVIDENCE, LANG-EDIT
LoadProfile: high (multi-layer inference)
PhaseTarget: {P2 by Sec2, P3 by Sec4 top band}
TransferTests: {explain with evidence; tone/purpose}
Sensors: {TR-comp, NIT-lite}
FailureModes: {OpinionAnswer, NoQuote}
RepairLoops: {quote→explain→link template}
FenceRules: enforce evidence template before grading (Λ control)
Node: SG×ENG×Z0×Learner×Node×LANG-EDIT
Prereq: [GRAM-CORE]
Outputs: LANG-SYNTH-SHORT
LoadProfile: high (error detection)
PhaseTarget: {P2 by Sec3}
TransferTests: {editing passage timed}
Sensors: {GR}
FailureModes: {RandomCorrections}
RepairLoops: {error-category map}
FenceRules: if RandomCorrections → truncate to 3 categories/week
Node: SG×ENG×Z0×Learner×Node×LANG-SYNTH-SHORT
Prereq: [COMP-SEC-INFERENCE, PARA-CORE]
Outputs: JC-SYN readiness
LoadProfile: high (compression)
PhaseTarget: {P2 by Sec4}
TransferTests: {summary; situational writing brief}
Sensors: {SR, CR}
FailureModes: {CopyTooMuch, MissKeyPoints}
RepairLoops: {highlight→group→compress protocol}
FenceRules: enforce 3-bucket grouping before writing
----------------------------------------
4.2 WRITING (NARRATIVE/DESCRIPTIVE + EARLY ARGUMENT)
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×WRIT-NARR-COHERENCE
Prereq: [COMP-DRAFT, COMP-PLAN-5BOX]
Outputs: WRIT-ARG-BASIC
LoadProfile: extreme
PhaseTarget: {P2 by Sec2, P3 by Sec4 top band}
TransferTests: {write unseen prompt timed}
Sensors: {NIT, CR, SII}
FailureModes: {NIT crossing: plot contradiction, POV/time jumps, causality breaks}
RepairLoops: {3-scene spine + cause-effect chain + rewrite weakest paragraph}
FenceRules:
- If NIT triggers → TRUNCATE story complexity → STITCH coherence first; ban extra twists
Node: SG×ENG×Z0×Learner×Node×WRIT-ARG-BASIC
Prereq: [PARA-CORE, COMP-SEC-EVIDENCE]
Outputs: JC-AQ readiness
LoadProfile: high (reasoning + structure)
PhaseTarget: {P2 by Sec4}
TransferTests: {claim→reason→example; rebuttal}
Sensors: {CR, NIT-arg}
FailureModes: {CircularReasoning, ExampleMismatch}
RepairLoops: {CER template + counterexample training}
FenceRules: if CircularReasoning persists → truncate to single-claim paragraph drills
========================================================
5) JC — NODE SET v0.1
========================================================
----------------------------------------
5.1 ARGUMENT / DISCOURSE / SYNTHESIS
----------------------------------------
Node: SG×ENG×Z0×Learner×Node×JC-AQ-THESIS
Prereq: [WRIT-ARG-BASIC]
Outputs: JC-AQ-REBUT, JC-DISC
LoadProfile: extreme (timed logic)
PhaseTarget: {P2 majority, P3 top band}
TransferTests: {thesis + 2 lines of reasoning in 10 min}
Sensors: {SR, CR, NIT-arg}
FailureModes: {ThesisVague, ReasoningLeak}
RepairLoops: {thesis sharpening + logic chain drills}
FenceRules: if ThesisVague → truncate scope; enforce “one clear claim”
Node: SG×ENG×Z0×Learner×Node×JC-SYN-COMPRESSION
Prereq: [LANG-SYNTH-SHORT]
Outputs: JC-EVID-CRIT
LoadProfile: extreme (compression + accuracy)
PhaseTarget: {P2 by JC2}
TransferTests: {summary within word limit}
Sensors: {SR, GR}
FailureModes: {Overrun, Distortion}
RepairLoops: {constraint-first drafting}
FenceRules: if Overrun repeated → truncate to outline-only drills
Node: SG×ENG×Z0×Learner×Node×JC-EVID-CRIT
Prereq: [COMP-SEC-INFERENCE]
Outputs: JC-DISC
LoadProfile: high
PhaseTarget: {P2 by JC2}
TransferTests: {evaluate evidence; bias; limitation}
Sensors: {CR}
FailureModes: {TrustEverything, NoCritique}
RepairLoops: {critique checklist}
FenceRules: enforce critique checklist
Node: SG×ENG×Z0×Learner×Node×JC-DISC
Prereq: [JC-AQ-THESIS, JC-EVID-CRIT]
Outputs: EXAM-P3 performance
LoadProfile: extreme
PhaseTarget: {P3 top band}
TransferTests: {full paper simulation}
Sensors: {TR, SII, NIT}
FailureModes: {TimeCrash, StructureCollapse}
RepairLoops: {section timing plan + weakest-section repair}
FenceRules: paper sims only after node stability (avoid premature overload)
========================================================
6) CORE EDGES (PREREQUISITE + TRANSFER) — v0.1
========================================================
Edge: SG×ENG×Z0×Operator×Edge×E-READ→VOC
From: READ-FLUENCY
To: VOC-ACTIVE-100
Weight: high
TypicalFailure: shallow exposure; no retrieval
RepairBridge: micro-dose reading + Fence Method retrieval
Edge: SG×ENG×Z0×Operator×Edge×E-VOC→SENT
From: VOC-ACTIVE-100
To: SENT-CORE
Weight: very high
TypicalFailure: knows word but cannot use
RepairBridge: word→phrase→sentence drills
Edge: SG×ENG×Z0×Operator×Edge×E-SENT→PARA
From: SENT-CORE
To: PARA-CORE
Weight: high
TypicalFailure: drift mid paragraph
RepairBridge: paragraph skeleton + checklist
Edge: SG×ENG×Z0×Operator×Edge×E-PARA→COMP
From: PARA-CORE
To: COMP-DRAFT
Weight: very high
TypicalFailure: incoherent essay; NIT crossing
RepairBridge: 5-Box plan + 3-scene spine + rewrite weakest paragraph
Edge: SG×ENG×Z0×Operator×Edge×E-INF→EVID
From: READ-COMP-INF
To: COMP-SEC-INFERENCE
Weight: high
TypicalFailure: opinion answers
RepairBridge: quote→explain→link template
Edge: SG×ENG×Z0×Operator×Edge×E-ARG→JC
From: WRIT-ARG-BASIC
To: JC-AQ-THESIS
Weight: very high
TypicalFailure: vague thesis
RepairBridge: thesis sharpening drills
========================================================
7) NIT SENSOR (NARRATIVE IRREVERSIBILITY THRESHOLD) — v0.1
========================================================
Definition:
NIT := point where output becomes internally inconsistent such that
the learner cannot self-correct even when prompted.
NIT Triggers (Primary→Secondary narrative):
- POV/time jumps without explanation
- causality breaks (event occurs without cause)
- character motive contradiction
- “and then” chain with no logic
- ending does not resolve initial problem
NIT Triggers (Argument):
- thesis shifts mid-essay
- circular reasoning
- evidence does not support claim
- rebuttal attacks wrong point
Fence Rule:
- If NIT triggers → TRUNCATE complexity (plot/claims)
- STITCH coherence first: 3-scene spine OR claim→reason→evidence
- Only after coherence stable → allow style upgrades
========================================================
8) CIVOS-HD SENSOR PACK (ENGLISH EDITION) — v0.1
========================================================
SensorPack: SG×ENG×Z0×Tutor×SensorPack×SENS-ENG-HD
Inputs:
- RR, GR, CR, SR, NIT, TR, SII, TTC, BN
Derived:
- DriftVelocity: d(RR)/dt, d(GR)/dt, d(CR)/dt
- Fence Timing: Θ, Λ (qualitative if needed)
Triggers:
- RR < 60% → stop adding new words; stitch retrieval
- GR rises under timing → shorten output; fix 3 error classes first
- CR unstable → paragraph skeleton drills
- NIT triggers → reduce complexity; rebuild coherence
- BN>3 with TTC<8w → truncate genre breadth; focus on exam-typical tasks
========================================================
9) FENCE RULES (GLOBAL STOP-LOSS) — v0.1
========================================================
FenceRules.Global:
- Meaning > Style. If meaning unstable, ban style upgrades.
- Coherence > Vocabulary fireworks. If coherence breaks, reduce word-level fancy.
- If NIT triggers twice in a week → mandatory coherence repair cycle.
- If TTC short and TR low → truncate to highest-yield nodes:
  {READ-COMP-INF, VOC-ACTIVE-100, GRAM-CORE, COMP-PLAN-5BOX, COMP-DRAFT}
========================================================
10) CH/ai SCHEDULER HOOK (ENGLISH) — v0.1
========================================================
CH/ai.Inputs:
- TTC, HoursPerWeek, NodePhaseScores, BN
- Weakest-link priority: NIT risk + RR risk + GR under time
- VolatilityIndex (school events)
CH/ai.Outputs:
- WeeklyRoutePlan:
  - 2× Retrieval Sessions (VOC)
  - 2× Sentence/Grammar Sessions
  - 2× Paragraph/Coherence Sessions
  - 1× Timed Writing Simulation (only if node stability achieved)
- DowngradeLogic:
  1) remove style tasks
  2) keep coherence + grammar + core vocab
  3) keep one timed simulation per week after stability
- ReTestSchedule:
  - RR test weekly
  - GR timed edit weekly
  - NIT check every writing attempt
========================================================
11) REQUIRED SHORT FAILURE MODE TRACE (CANONICAL)
========================================================
Trace:
Weak retrieval (RR low) → sentence instability (GR up) → paragraph drift (CR down) →
NIT crossing in composition → time crash → P2→P0 exam failure
Fence Repair:
Detect RR/GR/CR drift early → Truncate complexity → Stitch meaning+coherence →
timed ramps → maintain P2/P3 reliability
========================================================
END: SG-ENGLISH Master Graph v0.1
========================================================

WORDPRESS PASTE-READY PAGE STRUCTURE
H1: Singapore English Syllabus as a Lattice — SG-ENGLISH Master Graph v0.1
H2: Why a lattice beats a syllabus list (Detect→Predict→Repair)
H2: Phase (P0–P3) and Zoom (Z0–Z6) in Singapore English
H2: Primary Nodes: Reading → Vocabulary → Sentence → Paragraph → Composition
H2: Secondary Nodes: Inference → Evidence → Editing → Writing (Narrative + Argument)
H2: JC Nodes: Argument, Synthesis, Evidence Critique, Discourse Control
H2: Core Edges (Prerequisites + Transfer)
H2: NIT Sensor (Narrative Irreversibility Threshold)
H2: CivOS-HD Sensor Pack + Fence Rules (Stop-loss + Stitching)
H2: CH/ai Scheduler Hook (weekly route plan generator)
H2: Failure Mode Trace + Repair Trace