Slang Lineages & Emergent Branches

Status: Active documentation surface
Version: expanded 2026-08-05 (schema + matcher + confidence scoring + HTML + additional families + YTD backfill/backprop)
Purpose: Provide canonical visual and structural documentation for historical families of slang words and the processes by which new branches emerge. This layer supports Hyperlex analysis outputs, Abraxas Slang Module family-tree sections, and Orchestra symbolic mapping.

Interactive overview

For a dynamic radial map of all families and terms (click hubs, search leaves), open the Slang lineage map. Static Mermaid family trees remain below and under examples/slang-families/.

Why Lineages Matter

Slang is not a flat inventory of terms. It forms phylogenetic structures:

• Roots: older forms, often from specialized communities (criminal cant, occupational jargon, AAVE, military, betting syndicates, early internet/gaming).
• Trunk / Core family: stable semantic and social payload that persists across mutations.
• Branches: new senses, intensifiers, ironic inversions, platform-specific compressions, or hyperstitious loops.
• Emergent leaves: recent neologisms or eggcorn events that may stabilize or die.

Hyperlex treats these structures as first-class signals. detect_neologisms flags candidate branch points. trace_semantic_variation maps drift drivers. simulate_hyperstition_loop identifies self-reinforcing pathways that can accelerate a branch into cultural infrastructure.

A well-documented lineage turns a single observed term into a structured signal: it tells the system what family it belongs to, what payload it still carries, how far it has drifted, and whether it is currently actualizing.

Mutation Operators

These are the primary ways new branches form. Document them explicitly when reconstructing a family.

Operator	Description	Example
Extra-grammatical formation	Novel compounds, blends, or zero-derivations that violate ordinary morphology	“low block”, “false nine” style candidates
Sense extension / specialization	Existing term gains a tighter domain meaning	“steam” from general force → coordinated sharp line pressure
Irony inversion	Positive or neutral term flipped for status or critique	“based” → layered ironic uses; “aura” positive → “negative aura”
Platform compression	Extreme shortening or orthographic mutation optimized for a medium	“rekt”, “HODL”, vowel-dropped forms
Eggcorn / folk etymology	Mishearing or reanalysis that stabilizes as a new form	classic eggcorn events treated as symbolic adaptation
Cross-family borrowing	Term migrates and re-specializes in a new community	gaming “rekt” → crypto; WallStreetBets “diamond hands” → crypto
Hyperstition loop	Narrative about the term begins to produce the behavior the term describes	“steam” chase behavior reinforcing the line-move narrative

Documentation Method

1. Historical reconstruction — earliest attested uses (Green’s Dictionary of Slang, OED, domain glossaries, forum archives, arXiv cultural-transmission papers).
2. Semantic payload mapping — core emotional, identity, and routing functions that survive mutations. Ask: what social work does this family still do?
3. Branch points — mark documented mutations with operator type and approximate era.
4. Emergent monitoring — live signals from X, Reddit, Urban Dictionary, domain glossaries, and Hyperlex ingest feed candidate new leaves.
5. Visual encoding — Mermaid diagrams stored under examples/slang-families/.
6. Provenance tagging — every claim about a root or branch should be OBSERVED / INFERRED / SPECULATIVE where possible.
7. Registry + matcher — add the family to LINEAGE_REGISTRY in src/hyperlex/analysis/__init__.py so match_lineage() can attach it at runtime.

Documentation Template (copy for new families)

## Family: [Name]

**Domain**: [betting | crypto | kinship | internet-compression | ai-native | political-status | ...]
**Core payload**: [one-sentence identity / emotional / routing function]
**Root (OBSERVED)**: [earliest form + source + era]
**Trunk**: [stable core terms]
**Key branches**:
- [term] — [operator] — [era] — [payload shift]
**Current emergent leaves**: [list with confidence]
**Hyperstition risk**: [low/medium/high + brief mechanism]
**Diagram**: examples/slang-families/[name].mmd

Core Diagram Types

Type	Mermaid construct	Use
Family tree / mindmap	mindmap or flowchart TB	Historical root → core → branches → leaves
Emergence process	flowchart LR	Signal intake → variation → hyperstition → archive
Drift timeline	timeline or sequenced flowchart	Chronological mutation events
Network of related terms	graph	Cross-family borrowing and convergence

Integration Points

• Analysis module (zone_of_emergence): match_lineage() runs inside detect_memetic_patterns and attaches a lineage object under analysis when a match is found.
• Schema: schemas/lineage.v1.schema.json defines the attachment shape (including optional score_breakdown).
• Receipts: provenance can later include lineage_refs pointing to documented families.
• Abraxas Slang Emulation: the mandatory SLANG FAMILY TREE section in SIGNAL REPORTs is the live counterpart of these static diagrams.
• Orchestra: diagrams carry the orchestra-diagram.v1 header pattern already used in examples/hyperlex-symbolic/.

Live Lineage Attachment (current)

"analysis": {
  "neologisms": [...],
  "semantic_variation": {...},
  "lineage": {
    "family_id": "betting-sharp",
    "matched_terms": ["steam", "sharp"],
    "branch_operator": "sense_extension",
    "confidence": 0.72,
    "diagram_ref": "examples/slang-families/betting-sharp-family.mmd",
    "payload_note": "professional edge vs public money; line-physics signaling",
    "provenance": "INFERRED",
    "score_breakdown": {
      "n_hits": 2,
      "specificity": 0.41,
      "coverage": 0.22,
      "hit_bonus": 0.22,
      "density": 0.06,
      "raw": 0.72,
      "term_weights": {"steam": 0.37, "sharp": 0.37}
    }
  }
}

Confidence Scoring

compute_lineage_confidence(hits, family_terms, corpus) produces a deterministic score in [0, 0.98].

Components

Component	What it measures	Role
specificity	Average term-weight of the hits	Longer and multi-word terms (e.g. “diamond hands”, “aura farming”) are more distinctive than short common ones (“ape”, “mid”, “bro”)
coverage	len(hits) / len(family_terms)	Fraction of the family’s known vocabulary that appeared
hit_bonus	Diminishing returns per additional distinct hit	1st hit ≈ 0.12, 2nd ≈ 0.10, … floor ≈ 0.04; capped
density	Co-occurrence of ≥2 hits in a compact corpus	Multiple related terms close together is stronger evidence than scattered single hits

Term weight (specificity prior):

weight = min(0.75, 0.22 + 0.14 * n_words + 0.025 * min(len(term), 24))

Raw score:

raw = 0.18 + specificity * 0.38 + coverage * 0.22 + hit_bonus + density
confidence = min(0.98, max(0.0, raw))

Matching rules - Multi-word terms: substring match (already distinctive). - Single-word terms: word-boundary match (\bterm\b) to avoid false positives (“steam” inside “steamed”, “ape” inside “escape”).

Threshold: LINEAGE_CONFIDENCE_THRESHOLD = 0.42. Matches below this are discarded so weak single short-term hits do not attach a lineage.

The full breakdown is returned under score_breakdown for auditability and future calibration.

YTD backfill + lineage backpropagation

Hyperlex can backfill curated slang terms for prior months of the year and backpropagate lineage labels onto historical receipts without rewriting integrity.

Piece	Location	Role
Monthly packs	data/backfill/2026/YYYY-MM.json	Curated term seeds (OBSERVED / INFERRED / SPECULATIVE)
Loader / merge	hyperlex.analysis.backfill	Inventory + in-memory registry overlay
Rematch report	hyperlex.analysis.backprop	Re-run match_lineage on goldens/archive/local receipts
CLI	lineage-backfill, lineage-backprop	Operator surface

python3 scripts/hyperlex.py lineage-backfill --list --through 2026-08
python3 scripts/hyperlex.py lineage-backprop --from-golden --out out/backprop/report.json

Integrity rules

1. Historical receipt JSON and provenance.integrity are never rewritten by backprop.
2. Report schema: hyperlex.lineage_backprop.v1 (change classes: unchanged, gained, lost, reclassified, confidence_shift).
3. Brier remains null until settlement — backfill does not invent scores.
4. Optional: re-run archive-export for a new sanitized Pages snapshot after reviewing the report.

match_lineage(text, registry=...) accepts an overlay so backprop can use merged packs without permanently mutating the process-global LINEAGE_REGISTRY. Base registry still includes the main 2026 leaves so live analyze benefits immediately.

See data/backfill/2026/README.md for pack schema and month notes.
Visual timeline: examples/slang-families/ytd-2026-timeline.mmd.

Example Families Documented

See examples/slang-families/:

• betting-sharp-family.mmd + sharp-family-timeline.mmd — sharp/steam/square cluster and chronological drift.
• kinship-address.mmd — bro/sis/twin/unc lineage and platform acceleration.
• crypto-degen-family.mmd — HODL → diamond hands → ape/rekt/degen cluster.
• brainrot-aura-family.mmd — content-degradation + status-signaling (mid/cooked/aura/brainrot).
• ai-native-family.mmd — hallucinate → slop → clanker / agentic.
• political-status-family.mmd — based / redpilled / cope tribal-judgment lineage.
• emergence-process.mmd — abstract Hyperlex process that generates new branches.

HTML renderers: render-betting-sharp.html, render-ai-native.html, render-emergence.html.

Live Feed Process

1. Hyperlex (or Abraxas Slang Module) detects candidate terms via ingest + neologism pipeline.
2. match_lineage() scores against the static registry (seeded from the families above) using the confidence formula above.
3. Highest-confidence match that clears the threshold is attached under analysis.lineage.
4. If no match and confidence would be high for a novel cluster, a provisional leaf or family is proposed for human documentation.
5. Documentation is updated (diagram + markdown entry + registry entry) only after human review of provenance.
6. Future path: receipt histories → candidate diagram diffs → human approval gate.

Future Work

• Full schema validation of analysis.lineage inside validate_result.
• Automated diagram generation / diffing from receipt histories.
• Expanded cross-domain libraries (regional, sports beyond betting, finance subtypes).
• Brier-calibrated forecasts of branch survivability (using the confidence score as a prior).
• Richer interactive Orchestra-style HTML (node tooltips, flow highlighting).
• Learned term weights from historical receipt outcomes instead of the current heuristic.

References

• Green’s Dictionary of Slang (historical backbone)
• arXiv papers listed in references/arxiv_papers.md (semantic variation, cultural transmission, memetics, hyperstition)
• Hyperlex DESIGN principles (especially Real Over Synthetic, Provenance, ArXiv-Grounded, Lineage as First-Class Structure)
• Domain primary sources: BitcoinTalk archives, WallStreetBets, early gaming forums, Action Network glossary, Urban Dictionary attestations, Simon Willison / mainstream coverage of “slop” (2024)

Registry families (matcher seed)

betting-sharp · crypto-degen · ai-native · brainrot-aura · kinship-address · political-status · gaming-meta · workplace-corp