Liking cljdoc? Tell your friends :D

Unified Correction WFST Architecture

This document introduces the multi-tier Weighted Finite State Transducer (WFST) architecture for correction across written, spoken, and programming languages. The architecture includes the foundational three-tier WFST plus additional layers for dialogue context, LLM integration, and adaptive learning.

Sources:

  • liblevenshtein: /home/dylon/Workspace/f1r3fly.io/liblevenshtein-rust/
  • MORK: /home/dylon/Workspace/f1r3fly.io/MORK/
  • MeTTaTron: /home/dylon/Workspace/f1r3fly.io/MeTTa-Compiler/

Related Integration Docs:

Extended Architecture Docs:

Original WFST Documentation (detailed implementation specs):

  • WFST Architecture - Complete system design (~2400 lines)
  • CFG Grammar Correction - Error grammar formalism (~1900 lines)
  • Lattice Parsing - Earley parsing on lattices (~1050 lines)
  • Lattice Data Structures - Rust implementations (~550 lines)
  • NFA Phonetic Regex - Phonetic pattern matching
  • References - 35+ cited papers

Programming Language Correction (5-layer design with SMT repair):


Table of Contents

  1. Problem Statement
  2. Extended Architecture Overview
  3. Three-Tier WFST Core
  4. Dialogue Context Layer
  5. LLM Integration Layer
  6. Agent Learning Layer
  7. MORK Integration Phases
  8. Why Layered Correction?
  9. PathMap as Universal Storage
  10. Performance Considerations

Problem Statement

Error correction spans multiple domains with distinct requirements:

DomainError TypesCorrection Needs
Written TextTypos, spelling, grammarDictionary lookup, context
Spoken LanguagePhonetic confusion, homophonesPhoneme similarity, ASR lattices
Programming LanguagesSyntax errors, type mismatchesGrammar validation, semantic types

A unified architecture must handle all these while maintaining:

  • Efficiency: Real-time correction for interactive use
  • Accuracy: High precision without false corrections
  • Extensibility: Easy addition of new languages/domains

Extended Architecture Overview

The complete correction architecture extends the three-tier WFST core with additional layers for dialogue context, LLM integration, and adaptive learning. This multi-layer design enables:

  • Conversational correction: Understanding context across dialogue turns
  • LLM agent integration: Pre/post-processing for language model interactions
  • Personalization: Learning from user feedback and preferences
┌─────────────────────────────────────────────────────────────────────────┐
│          EXTENDED CORRECTION ARCHITECTURE (Full Stack)                   │
├─────────────────────────────────────────────────────────────────────────┤
│                                                                          │
│  ┌────────────────────────────────────────────────────────────────────┐ │
│  │                    DIALOGUE CONTEXT LAYER                          │ │
│  │  Turn History │ Entity Registry │ Topic Graph │ Speaker Models     │ │
│  │  [Discourse semantics, coreference resolution, topic tracking]     │ │
│  │  See: ../dialogue/README.md                                        │ │
│  └────────────────────────────────────────────────────────────────────┘ │
│                              │                                          │
│  ┌───────────────────────────┼────────────────────────────────────────┐ │
│  │                           ▼                                        │ │
│  │               THREE-TIER WFST CORE                                 │ │
│  │  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐             │ │
│  │  │ Tier 1:      │→ │ Tier 2:      │→ │ Tier 3:      │             │ │
│  │  │ Lexical      │  │ Syntactic    │  │ Semantic     │             │ │
│  │  │ (libleven.)  │  │ (MORK/CFG)   │  │ (MeTTaIL)    │             │ │
│  │  └──────────────┘  └──────────────┘  └──────────────┘             │ │
│  │  [Edit distance, phonetic rules, grammar validation, type checking]│ │
│  │  See: #three-tier-wfst-core below                                  │ │
│  └───────────────────────────────────────────────────────────────────┘ │
│                              │                                          │
│  ┌───────────────────────────┼────────────────────────────────────────┐ │
│  │                           ▼                                        │ │
│  │               PRAGMATIC REASONING LAYER                            │ │
│  │  Speech Act Classifier │ Implicature Resolver │ Relevance Ranker   │ │
│  │  [Intent detection, indirect speech acts, contextual relevance]    │ │
│  │  See: ../dialogue/04-pragmatic-reasoning.md                        │ │
│  └───────────────────────────────────────────────────────────────────┘ │
│                              │                                          │
│  ┌───────────────────────────┼────────────────────────────────────────┐ │
│  │                           ▼                                        │ │
│  │               LLM INTEGRATION LAYER                                │ │
│  │  ┌────────────────────────────────────────────────────────────┐   │ │
│  │  │ PROMPT PREPROCESSING                                       │   │ │
│  │  │ Correction → Coreference → Context Injection → RAG         │   │ │
│  │  └────────────────────────────┬───────────────────────────────┘   │ │
│  │                               ▼                                    │ │
│  │                       ┌──────────────┐                            │ │
│  │                       │   LLM API    │                            │ │
│  │                       └──────┬───────┘                            │ │
│  │                              ▼                                     │ │
│  │  ┌────────────────────────────────────────────────────────────┐   │ │
│  │  │ RESPONSE POSTPROCESSING                                    │   │ │
│  │  │ Coherence Check → Fact Verification → Hallucination Detect │   │ │
│  │  └────────────────────────────────────────────────────────────┘   │ │
│  │  See: ../llm-integration/README.md                                │ │
│  └────────────────────────────────────────────────────────────────────┘ │
│                              │                                          │
│  ┌───────────────────────────┼────────────────────────────────────────┐ │
│  │                           ▼                                        │ │
│  │               AGENT LEARNING LAYER                                 │ │
│  │  Feedback Collection │ Pattern Learning │ User Preferences         │ │
│  │  Online Learning │ Threshold Adaptation │ Model Versioning         │ │
│  │  [Adaptive correction weights, personalized dictionaries]          │ │
│  │  See: ../agent-learning/README.md                                  │ │
│  └────────────────────────────────────────────────────────────────────┘ │
│                                                                          │
└──────────────────────────────────────────────────────────────────────────┘

Layer Summary

LayerComponentsPurpose
Dialogue ContextTurn History, Entity Registry, Topic GraphMulti-turn conversation tracking
WFST CoreLexical, Syntactic, Semantic TiersFundamental correction pipeline
SimplificationAnalysis, Rules, Strategy, VerificationPost-correction source optimization
PragmaticSpeech Acts, Implicatures, RelevanceIntent understanding
LLM IntegrationPreprocessing, PostprocessingLLM agent support
Agent LearningFeedback, Patterns, PreferencesAdaptive personalization

Three-Tier WFST Core

The foundational correction system uses three progressively refined tiers:

┌─────────────────────────────────────────────────────────────────────┐
│           UNIFIED CORRECTION WFST ARCHITECTURE                       │
├─────────────────────────────────────────────────────────────────────┤
│                                                                      │
│  INPUT: Erroneous text (written/spoken/code)                        │
│                                                                      │
│  ┌──────────────────────────────────────────────────────────────┐   │
│  │                    Tier 1: Lexical Correction                 │   │
│  │                       (liblevenshtein)                        │   │
│  │  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐          │   │
│  │  │ Edit Dist.  │  │  Phonetic   │  │   Custom    │          │   │
│  │  │ Automata    │  │   Rules     │  │   Weights   │          │   │
│  │  └──────┬──────┘  └──────┬──────┘  └──────┬──────┘          │   │
│  │         └─────────────────┼─────────────────┘                │   │
│  │                           ▼                                   │   │
│  │              ┌────────────────────────┐                      │   │
│  │              │   Candidate Lattice    │                      │   │
│  │              └───────────┬────────────┘                      │   │
│  └──────────────────────────┼───────────────────────────────────┘   │
│                             ▼                                        │
│  ┌──────────────────────────────────────────────────────────────┐   │
│  │                  Tier 2: Syntactic Validation                 │   │
│  │                     (CFG + MORK/PathMap)                      │   │
│  │  ┌─────────────────────────────────────────────────────────┐ │   │
│  │  │                     MORK Space                          │ │   │
│  │  │  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐     │ │   │
│  │  │  │  Grammar    │  │  Pattern    │  │   Bloom +   │     │ │   │
│  │  │  │  Rules      │  │  Matching   │  │   LRU       │     │ │   │
│  │  │  └──────┬──────┘  └──────┬──────┘  └──────┬──────┘     │ │   │
│  │  │         └─────────────────┼─────────────────┘           │ │   │
│  │  │                           ▼                             │ │   │
│  │  │               ┌─────────────────────┐                   │ │   │
│  │  │               │      PathMap        │                   │ │   │
│  │  │               │  (Shared Storage)   │                   │ │   │
│  │  │               └──────────┬──────────┘                   │ │   │
│  │  └──────────────────────────┼──────────────────────────────┘ │   │
│  │                             ▼                                 │   │
│  │              ┌────────────────────────┐                      │   │
│  │              │  Syntactically Valid   │                      │   │
│  │              │     Candidates         │                      │   │
│  │              └───────────┬────────────┘                      │   │
│  └──────────────────────────┼───────────────────────────────────┘   │
│                             ▼                                        │
│  ┌──────────────────────────────────────────────────────────────┐   │
│  │                Tier 3: Semantic Type Checking                 │   │
│  │                (MeTTaIL / MeTTaTron / Rholang)                │   │
│  │  ┌─────────────────────────────────────────────────────────┐ │   │
│  │  │                    MeTTaTron                            │ │   │
│  │  │  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐     │ │   │
│  │  │  │   MeTTa     │  │   Type      │  │  Behavioral │     │ │   │
│  │  │  │   Atomspace │  │   Checking  │  │   Predicates│     │ │   │
│  │  │  └──────┬──────┘  └──────┬──────┘  └──────┬──────┘     │ │   │
│  │  │         └─────────────────┼─────────────────┘           │ │   │
│  │  │                           ▼                             │ │   │
│  │  │        ┌──────────────────────────────────┐             │ │   │
│  │  │        │ OSLF Predicate Evaluation        │             │ │   │
│  │  │        │ (structural + behavioral types)  │             │ │   │
│  │  │        └───────────────┬──────────────────┘             │ │   │
│  │  └────────────────────────┼────────────────────────────────┘ │   │
│  │                           │                                   │   │
│  │  ┌────────────────────────┼────────────────────────────────┐ │   │
│  │  │                   Rholang Bridge                        │ │   │
│  │  │  PathMap <-> MeTTa State <-> Rholang Par                │ │   │
│  │  │  (Enables cross-language semantic checking)              │ │   │
│  │  └────────────────────────┼────────────────────────────────┘ │   │
│  │                           ▼                                   │   │
│  │              ┌────────────────────────┐                      │   │
│  │              │  Semantically Valid    │                      │   │
│  │              │     Corrections        │                      │   │
│  │              └───────────┬────────────┘                      │   │
│  └──────────────────────────┼───────────────────────────────────┘   │
│                             ▼                                        │
│  OUTPUT: Ranked corrections with confidence scores                   │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

Tier Summary

TierComponentPurposeSpeed
1liblevenshteinLexical candidates via edit distanceFastest
2MORK/PathMapSyntactic filtering via CFGFast
3MeTTaIL/RholangSemantic type checkingThorough

Dialogue Context Layer

The dialogue context layer extends correction capabilities for multi-turn conversations, enabling context-aware corrections that consider the full discourse history.

Full documentation: Dialogue Context Documentation

Components

ComponentPurposePathMap Key
Turn TrackerConversation history with sliding window/dialogue/{id}/turn/
Entity RegistryCross-turn entity tracking and coreference/dialogue/{id}/entity/
Topic GraphDiscourse structure and topic continuity/dialogue/{id}/topic/
Speaker ModelsPer-participant vocabulary and style/dialogue/{id}/speaker/

Key Capabilities

  1. Coreference Resolution: Resolves pronouns and references across turns

    • Pronoun resolution: "it" → "the document"
    • Definite description binding: "the file" → specific file entity
    • See: Coreference Resolution
  2. Discourse Coherence: Validates corrections maintain conversation flow

    • Coherence relations (Elaboration, Question-Answer, Contrast)
    • Topic continuity checking
    • See: Discourse Semantics
  3. Topic Management: Tracks and validates topic shifts

    • Topic extraction and clustering
    • Drift detection and validation
    • See: Topic Management

Integration with WFST Core

Dialogue Context → WFST Core
─────────────────────────────
• Entity salience affects candidate ranking
• Topic keywords influence lexical tier
• Speaker vocabulary personalizes dictionary
• Discourse coherence validates semantic tier

LLM Integration Layer

The LLM integration layer provides preprocessing and postprocessing for language model interactions, ensuring corrected input and validated output.

Full documentation: LLM Integration Documentation

Preprocessing Pipeline

Transforms user input before LLM processing:

User Input → Correction → Coreference → Context Injection → RAG → LLM Prompt
StageFunctionDocumentation
CorrectionThree-tier WFST fixes errorsThis document
CoreferenceResolves references using dialogue context02-coreference-resolution.md
Context InjectionFormats dialogue history for prompt04-context-injection.md
RAGRetrieves relevant knowledge04-context-injection.md

Documentation: Prompt Preprocessing

Postprocessing Pipeline

Validates and corrects LLM responses:

LLM Response → Coherence → Fact Check → Hallucination → Correction → Final Output
StageFunctionDocumentation
Coherence CheckValidates response addresses query02-output-postprocessing.md
Fact VerificationChecks factual claims against knowledge03-hallucination-detection.md
Hallucination DetectionIdentifies fabricated content03-hallucination-detection.md
CorrectionThree-tier WFST fixes errorsThis document

Documentation: Output Postprocessing

Hallucination Types

TypeDetection MethodExample
Fabricated FactKnowledge base mismatchInvented statistics
Nonexistent EntityEntity registry lookupMade-up person names
Temporal ErrorTimeline validationWrong date claims
ContradictionDialogue consistencyConflicting statements

Documentation: Hallucination Detection


Agent Learning Layer

The agent learning layer provides adaptive correction through feedback collection, pattern learning, and online weight updates.

Full documentation: Agent Learning Documentation

Components

ComponentPurposeDocumentation
Feedback CollectionCapture user responses to corrections01-feedback-collection.md
Pattern LearningExtract error patterns from feedback02-pattern-learning.md
User PreferencesModel individual user characteristics03-user-preferences.md
Online LearningIncremental weight and threshold updates04-online-learning.md

Feedback Flow

User Action → Signal Detection → Normalization → Learning Update
───────────────────────────────────────────────────────────────
Accept (fast)     → Strong positive   → +0.8 to +1.0
Accept (slow)     → Weak positive     → +0.3 to +0.5
Modify            → Correction signal → Pattern extraction
Reject            → Negative signal   → -0.8 to -1.0
Ignore            → Weak negative     → -0.1 to -0.3

Learned Adaptations

AdaptationScopeEffect
Edit WeightsGlobal/UserCharacter-level substitution costs
Feature WeightsGlobal/UserRanking factor importance
ThresholdsUser/DomainCorrection confidence cutoffs
VocabularyUserPersonal dictionary additions
PatternsGlobalRecognized error→correction pairs

PathMap Storage

/learning/
    /patterns/                 # Learned error patterns
    /user/{user_id}/          # Per-user profiles
        /vocabulary/          # Personal dictionary
        /weights/             # Personalized weights
        /thresholds/          # Confidence thresholds
    /models/                  # Version-controlled models
        /current/             # Active model
        /checkpoints/         # Historical snapshots

MORK Integration Phases

The three-tier architecture is implemented through four progressive phases, each building on the previous. See the MORK Integration Overview for complete implementation details.

┌─────────────────────────────────────────────────────────────────────┐
│                    MORK Integration Phases                           │
├─────────────────────────────────────────────────────────────────────┤
│                                                                      │
│  Phase A: FuzzySource Trait                                         │
│  ┌─────────────────────────────────────────────────────────────┐   │
│  │  • Trait abstraction for fuzzy dictionary backends          │   │
│  │  • PathMap + DAWG + DoubleArrayTrie implementations         │   │
│  │  • Integration point: liblevenshtein → MORK                 │   │
│  └─────────────────────────────────────────────────────────────┘   │
│                              ↓                                       │
│  Phase B: Lattice Infrastructure                                    │
│  ┌─────────────────────────────────────────────────────────────┐   │
│  │  • Weighted DAG for multi-candidate representation          │   │
│  │  • K-best path extraction (Dijkstra-based)                  │   │
│  │  • LatticeZipper for MORK ProductZipper integration         │   │
│  └─────────────────────────────────────────────────────────────┘   │
│                              ↓                                       │
│  Phase C: WFST Composition                                          │
│  ┌─────────────────────────────────────────────────────────────┐   │
│  │  • Semiring weights (Tropical, Log, Probability)            │   │
│  │  • Phonetic NFA via Thompson's construction                 │   │
│  │  • FST ∘ FST ∘ Trie composition operators                   │   │
│  └─────────────────────────────────────────────────────────────┘   │
│                              ↓                                       │
│  Phase D: Grammar Correction                                        │
│  ┌─────────────────────────────────────────────────────────────┐   │
│  │  • CFG rules as pattern/template pairs                      │   │
│  │  • MORK match2() for structural matching                    │   │
│  │  • query_multi_i() for O(K×N) lattice processing            │   │
│  └─────────────────────────────────────────────────────────────┘   │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

Phase A: FuzzySource Trait

Documentation: FuzzySource Implementation

The FuzzySource trait provides a unified interface for fuzzy dictionary lookups across different storage backends:

/// Unified trait for fuzzy dictionary sources.
pub trait FuzzySource {
    /// Query with fuzzy matching up to max_distance.
    fn fuzzy_lookup(&self, query: &[u8], max_distance: u8)
        -> impl Iterator<Item = (Vec<u8>, u8)>;
}

Implementations:

  • PathMap: Trie-based storage with zipper navigation
  • DynamicDawg / DynamicDawgChar: SIMD-optimized for runtime updates
  • DoubleArrayTrie / DoubleArrayTrieChar: Optimized for static dictionaries

Integration Point: Tier 1 (Lexical Correction) uses FuzzySource for candidate generation.

Phase B: Lattice Infrastructure

Documentation: Lattice Integration

Lattices represent the space of correction candidates as weighted directed acyclic graphs:

Query Term: "teh"
    │
    ▼
Transducer::query_lattice()
    │
    │ Builds DAG of candidates with weighted edges
    ▼
Lattice { nodes, edges, vocab }
    │
    ▼
LatticeZipper (MORK adapter)
    │
    │ Iterates paths by total weight
    ▼
ProductZipper → Unification → Ranked Results

Key Components:

  • Lattice: Core DAG structure with vocabulary deduplication
  • LatticeBuilder: Incremental construction API
  • PathIterator / k_best(): Path extraction algorithms
  • LatticeZipper: Adapter for MORK's ProductZipper

Integration Point: Bridge between Tier 1 and Tier 2.

Phase C: WFST Composition

Documentation: WFST Composition

Full Weighted Finite State Transducer infrastructure with phonetic NFA composition:

Query Pattern: "(ph|f)(o|oa)(n|ne)"
    │
    ▼
PhoneticNfa::compile()      ← Thompson's construction
    │
    ▼
ComposedAutomaton::new(phonetic_nfa, levenshtein, dictionary)
    │
    │ FST ∘ FST ∘ Trie composition
    ▼
Lattice with phonetic-weighted edges

Key Concepts:

Semiring$\oplus$ (combine)$\otimes$ (extend)Use Case
Tropicalmin+Shortest path (Viterbi)
Loglog-sum-exp+Probabilistic (forward-backward)
Probability+×Raw probabilities

Integration Point: Tier 1 phonetic expansion before Tier 2 filtering.

Phase D: Grammar Correction

Documentation: Grammar Correction

CFG-based error correction using MORK's pattern matching as the rule engine:

; CFG Rule: Subject-Verb Agreement Error
Pattern:  (s (np ?Subj :number singular) (vp (v ?V :number plural) ?Rest))
Template: (s (np ?Subj :number singular) (vp (v (singularize ?V)) ?Rest))
Cost: 1.0

Key MORK Functions:

FunctionLocationPurpose
match2()expr/src/lib.rs:921Recursive structural matching
unify()expr/src/lib.rs:1849Variable binding + constraints
query_multi_i()kernel/src/space.rs:992$\mathcal{O}(K\times N)$ lattice queries
transform_multi_multi_()kernel/src/space.rs:1221Pattern→template application

Integration Point: Tier 2 (Syntactic Validation) rule engine.

Phase Integration Summary

PhaseTierPrimary FunctionOutput
A1Fuzzy lookupRaw candidates
B1→2Lattice constructionWeighted DAG
C1Phonetic expansionExpanded candidates
D2Grammar filteringValid corrections

Why Layered Correction?

Progressive Refinement

Each tier reduces the candidate set before the next:

Input Error: "teh" in "teh cat sat"
    │
    ▼ Tier 1 (Lexical)
Candidates: [the, tea, ten, tee, tech, ...]  (~100 candidates)
    │
    ▼ Tier 2 (Syntactic)
Valid in context: [the, tea]  (grammar allows determiner or noun)
    │
    ▼ Tier 3 (Semantic)
Best correction: "the"  (matches "cat sat" semantic context)

Computational Efficiency

TierComplexityCandidates
1$\mathcal{O}(n \times d)$Generate many
2$\mathcal{O}(n \times g)$Filter structurally
3$\mathcal{O}(n \times t)$Verify semantically

Where:

  • n = number of candidates
  • d = edit distance bound
  • g = grammar size
  • t = type checking cost

By filtering at each tier, expensive semantic checks only run on valid candidates.

Separation of Concerns

Each tier has distinct expertise:

TierKnowledge Required
1Character/phoneme similarity
2Language grammar
3Type system, domain semantics

PathMap as Universal Storage

PathMap serves as the shared storage layer across all tiers:

┌─────────────────────────────────────────────────────────────────┐
│                    PathMap Integration                           │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────────────┐                                            │
│  │  liblevenshtein │                                            │
│  │  Dictionary     │──────┐                                     │
│  └─────────────────┘      │                                     │
│                           │                                     │
│  ┌─────────────────┐      │      ┌─────────────────────────┐   │
│  │  MORK Grammar   │──────┼─────>│       PathMap           │   │
│  │  Rules          │      │      │  (Trie-based Storage)   │   │
│  └─────────────────┘      │      └─────────────────────────┘   │
│                           │                 │                   │
│  ┌─────────────────┐      │                 │                   │
│  │  MeTTa Type     │──────┘                 ▼                   │
│  │  Predicates     │           ┌────────────────────────────┐  │
│  └─────────────────┘           │  Shared Query Interface     │  │
│                                │  - Pattern matching          │  │
│                                │  - Fuzzy lookup              │  │
│                                │  - Type queries              │  │
│                                └────────────────────────────┘  │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

Benefits

  1. No Serialization Overhead: All tiers operate on same data format
  2. Shared Indexing: Pattern matching works uniformly
  3. Cross-Tier Queries: Type predicates can reference grammar rules
  4. Merkleization: Content-addressed caching across tiers

Performance Considerations

Tier 1 Optimizations

  • SIMD: DynamicDawg uses SIMD for parallel character comparison
  • Bloom Filter: Fast negative lookups before trie traversal
  • Lazy Iteration: Candidates generated on-demand

Tier 2 Optimizations

  • Lattice Parsing: 3-10x speedup over exhaustive enumeration
  • LRU Cache: Hot grammar rules cached
  • Incremental Parsing: Reuse partial parses for nearby errors

Tier 3 Optimizations

  • Predicate Caching: Common type queries cached
  • Lazy Evaluation: Type checking on-demand
  • Parallel Checking: Independent candidates checked in parallel

Memory Budget

ComponentTypical Size
Dictionary (English)50-100 MB
Grammar (Programming Language)10-50 MB
Type Predicates5-20 MB
Working Set (LRU)10-50 MB

Summary

The extended correction architecture provides:

  1. Comprehensive Correction: Lexical, syntactic, and semantic (three-tier WFST core)
  2. Conversational Support: Multi-turn dialogue context and coreference resolution
  3. LLM Integration: Pre/post-processing for language model agent interactions
  4. Adaptive Learning: Feedback-driven personalization and online weight updates
  5. Efficient Filtering: Each tier reduces candidates before expensive processing
  6. Unified Storage: PathMap as shared layer across all components

Core Integration Points

FromToInterface
liblevenshteinPathMapFuzzySource trait
MORKPathMapNative storage backend
MeTTaTronPathMapType predicate storage
RholangPathMapPar conversion
Dialogue ContextWFST CoreEntity salience, speaker vocab
LLM LayerWFST CorePre/post-processing pipeline
Agent LearningAll LayersAdaptive weights and thresholds

Layer Dependencies

Dialogue Context Layer
        │
        ▼
Three-Tier WFST Core ←────────────────────────────┐
        │                                          │
        ▼                                          │
Pragmatic Reasoning Layer                          │
        │                                          │
        ▼                                          │
LLM Integration Layer                              │
        │                                          │
        ▼                                          │
Agent Learning Layer ──────────────────────────────┘
        │                  (feedback loop)
        ▼
   PathMap Storage

References

WFST Core Documentation

Extended Layer Documentation

Integration Documentation

Can you improve this documentation?Edit on GitHub

cljdoc builds & hosts documentation for Clojure/Script libraries

Keyboard shortcuts
Ctrl+kJump to recent docs
Move to previous article
Move to next article
Ctrl+/Jump to the search field
× close