Rules System - mySpellChecker

mySpellChecker uses YAML configuration files to define linguistic rules for Myanmar language. These files are located in src/myspellchecker/rules/ and can be customized for specific use cases.

Rule Files Overview

File	Purpose
`particles.yaml`	Linguistic particles with POS tags
`typo_corrections.yaml`	Common typo patterns
`morphology.yaml`	Suffix/prefix patterns
`morphotactics.yaml`	Morpheme combination rules
`grammar_rules.yaml`	Grammar validation rules
`aspects.yaml`	Verb aspect markers
`compounds.yaml`	Compound word patterns
`classifiers.yaml`	Numeral classifiers
`negation.yaml`	Negation patterns
`register.yaml`	Formal/colloquial mappings
`tone_rules.yaml`	Tone mark rules
`homophones.yaml`	Homophone pairs
`orthographic_corrections.yaml`	Orthographic correction patterns
`rerank_rules.yaml`	Suggestion re-ranking rules
`ambiguous_words.yaml`	Multi-POS words
`pos_inference.yaml`	POS inference patterns
`pronouns.yaml`	Pronoun definitions
`collocations.yaml`	Collocation error detection (wrong word near context words)
`compound_confusion.yaml`	Compound word confusion patterns (ha-htoe, aspiration, consonant, suffix)
`confusable_pairs.yaml`	Curated confusable word pairs for real-word error detection
`confusion_matrix.yaml`	Data-driven character substitution costs from corpus analysis
`corruption_weights.yaml`	Synthetic error generation weights for training
`detector_confidences.yaml`	Confidence scores for text-level detectors by error type
`homophone_confusion.yaml`	Context-dependent homophone disambiguation rules
`medial_confusion.yaml`	Medial ya-pin / ya-yit confusion correction patterns
`medial_swap_pairs.yaml`	Medial consonant swap pairs for candidate generation
`semantic_rules.yaml`	Semantic validation (agent implausibility, classifier agreement)
`stacking_pairs.yaml`	Valid consonant stacking pairs via virama (U+1039)
`tense_markers.yaml`	Temporal adverb and sentence-final particle tense mismatch rules

File Structure

All rule files follow a common structure:

version: "1.0.0"
category: "category_name"
description: "Description of the rule file"

metadata:
  created_date: "2025-12-30"
  last_updated: "2025-12-30"
  total_entries: 100
  source: "Source description"

# Main content section
rules:
  - ...

Particles (particles.yaml)

Defines Myanmar linguistic particles organized by syntactic function.

Structure

particles:
  verbs:
    tense:
      - particle: "ခဲ့"
        pos_tag: "P_PAST"
        type: "past_tense"
        meaning: "Past tense marker"
        formality: "neutral"
        examples:
          - correct: "သွားခဲ့တယ်"
            translation: "went"
        confidence: 0.98

    aspect:
      - particle: "နေ"
        pos_tag: "P_PROG"
        type: "progressive"
        meaning: "Progressive aspect"
        formality: "neutral"
        confidence: 0.98

POS Tags

Tag	Description	Example
`P_PAST`	Past tense	ခဲ့
`P_FUT`	Future tense	မယ်, မည်
`P_PROG`	Progressive	နေ
`P_PERF`	Perfective	ပြီ
`P_SUBJ`	Subject marker	က
`P_OBJ`	Object marker	ကို
`P_LOC`	Locative	မှာ, တွင်
`P_SENT`	Sentence ending	တယ်, သည်
`P_POSS`	Possessive	ရဲ့, ၏

Formality Levels

colloquial - Spoken/informal
neutral - Both formal and informal
formal - Written/formal
polite - Respectful register
literary - Literary style

Typo Corrections (typo_corrections.yaml)

Defines common Myanmar typo patterns with corrections.

Structure

corrections:
  particles:
    - incorrect: "မာ"
      correct: "မှာ"
      error_type: "missing_ha_htoe"
      context: "after_noun"
      excluded_pos: ["ADJ"]
      meaning: "Location particle"
      confidence: 0.92
      examples:
        incorrect: "အိမ်မာ ရှိတယ်"
        correct: "အိမ်မှာ ရှိတယ်"

  medial_confusions:
    - incorrect: "ကျောင်း"
      correct: "ကြောင်း"
      error_type: "ya_pin_ra_yit"
      context: "after_verb"
      pos_constraint:
        preceding: ["V"]
      confidence: 0.95

Error Types

Type	Description
`missing_ha_htoe`	Missing ှ modifier
`character_confusion`	Similar looking characters
`ya_pin_ra_yit`	ျ vs ြ confusion
`missing_asat`	Missing ် marker
`tone_mark_error`	Wrong or missing tone mark
`visual_similar`	OCR-type errors

Context Types

after_noun - Follows a noun
after_verb - Follows a verb
context_dependent - Requires context analysis
standalone - Independent of context

Morphology (morphology.yaml)

Defines suffix and prefix patterns for POS inference.

Structure

suffixes:
  verb_suffixes:
    - suffix: "ခဲ့"
      pos: "V"
      meaning: "past tense"
      confidence: 0.9

    - suffix: "သည်"
      pos: "P_SENT"
      meaning: "formal sentence ending"
      confidence: 0.95

  noun_suffixes:
    - suffix: "များ"
      pos: "N"
      meaning: "plural"
      confidence: 0.9

  adverb_suffixes:
    - suffix: "စွာ"
      pos: "ADV"
      meaning: "manner"
      confidence: 0.85

Aspects (aspects.yaml)

Defines verb aspect markers.

Structure

markers:
  - marker: "ပြီ"
    category: "completion"
    description: "Action completed"
    can_combine: false
    register: "neutral"
    is_final: true

  - marker: "နေ"
    category: "progressive"
    description: "Ongoing action"
    can_combine: true
    register: "neutral"
    is_final: false

combinations:
  - sequence: ["ပြီး", "သွား"]
    description: "Completed and went"

invalid_sequences:
  - sequence: ["ပြီ", "ပြီ"]
    reason: "Duplicate completion marker"

typos:
  - incorrect: "ပရီ"
    correct: "ပြီ"

Classifiers (classifiers.yaml)

Defines numeral classifiers for counting.

Structure

classifiers:
  people:
    - word: "ယောက်"
      description: "For people"
      examples: ["လူ", "ကလေး", "လူကြီး"]

  animals:
    - word: "ကောင်"
      description: "For animals"
      examples: ["ခွေး", "ကြောင်", "ငါး"]

  flat_objects:
    - word: "ရွက်"
      description: "For flat objects"
      examples: ["စာရွက်", "အရွက်"]

  round_objects:
    - word: "လုံး"
      description: "For round objects"
      examples: ["ပန်းသီး", "ဘောလုံး"]

Register (register.yaml)

Maps formal and colloquial equivalents.

Structure

register_pairs:
  - formal: "သည်"
    colloquial: "တယ်"
    category: "sentence_ending"

  - formal: "၏"
    colloquial: "ရဲ့"
    category: "possessive"

  - formal: "တွင်"
    colloquial: "မှာ"
    category: "locative"

formal_words:
  - "သည်"
  - "၏"
  - "နှင့်"

colloquial_words:
  - "တယ်"
  - "ရဲ့"
  - "နဲ့"

Negation (negation.yaml)

Defines negation patterns.

Structure

prefix: "မ"

endings:
  ဘူး:
    type: "standard_negative"
    description: "Colloquial negative ending"
    register: "colloquial"

  ပါဘူး:
    type: "polite_negative"
    description: "Polite negative ending"
    register: "polite"

  နဲ့:
    type: "prohibition"
    description: "Don't! (prohibition)"
    register: "colloquial"

  ပါ:
    type: "formal_negative"
    description: "Formal negative ending"
    register: "formal"

typo_map:
  ဘူ: "ဘူး"
  ဘုး: "ဘူး"

auxiliaries:
  ချင်:
    meaning: "want to"
  နိုင်:
    meaning: "can"
  ရ:
    meaning: "able to"

Homophones (homophones.yaml)

Defines homophone pairs for context checking.

Structure

homophones:
  # Simple map: word → list of homophones
  "ကား": ["ကာ"]            # car vs protect/shield
  "ကာ": ["ကား"]
  "ကျောင်း": ["ကြောင်း"]  # school vs reason
  "ကြောင်း": ["ကျောင်း"]
  "ကံ": ["ကန်", "ကင်"]    # luck vs kick vs (rare)
  "ကန်": ["ကံ", "ကင်"]

Context disambiguation is handled automatically via N-gram probabilities at the strategy level, so no per-entry disambiguation context is needed in the YAML.

Compounds (compounds.yaml)

Defines compound word formations.

Structure

prefixes:
  - prefix: "အ"
    type: "nominalization"
    description: "Noun-forming prefix"

suffixes:
  - suffix: "သူ"
    type: "agent"
    description: "Person who does X"

noun_compounds:
  - components: ["ပန်း", "ခြံ"]
    compound: "ပန်းခြံ"
    meaning: "flower garden"

verb_compounds:
  - components: ["စား", "သောက်"]
    compound: "စားသောက်"
    meaning: "dine/eat and drink"

reduplication:
  - base: "ဖြေး"
    reduplicated: "ဖြေးဖြေး"
    meaning: "slowly"

Custom Configuration

Loading Custom Rules

from myspellchecker.grammar.config import GrammarRuleConfig

# Load from custom YAML file path
config = GrammarRuleConfig(config_path="/path/to/custom/grammar_rules.yaml")

# Access rule data
particles = config.particle_tags
morphology = config.morphology_config
aspects = config.aspects_config

Extending Rules

Add custom entries by creating additional YAML files:

# custom_particles.yaml
version: "1.0.0"
category: "particles"

particles:
  custom:
    - particle: "မိ"
      pos_tag: "P_CUSTOM"
      type: "custom_type"
      meaning: "Custom particle"
      confidence: 0.80

Schema Validation

Rule files are validated against JSON schemas in src/myspellchecker/schemas/. There is one schema file per YAML rule file (30 total, including _common.schema.json for shared definitions):

_common.schema.json, shared definitions referenced by other schemas
ambiguous_words.schema.json
aspects.schema.json
classifiers.schema.json
collocations.schema.json
compound_confusion.schema.json
compounds.schema.json
confusable_pairs.schema.json
confusion_matrix.schema.json
corruption_weights.schema.json
detector_confidences.schema.json
grammar_rules.schema.json
homophone_confusion.schema.json
homophones.schema.json
medial_confusion.schema.json
medial_swap_pairs.schema.json
morphology.schema.json
morphotactics.schema.json
negation.schema.json
orthographic_corrections.schema.json
particles.schema.json
pos_inference.schema.json
pronouns.schema.json
register.schema.json
rerank_rules.schema.json
semantic_rules.schema.json
stacking_pairs.schema.json
tense_markers.schema.json
tone_rules.schema.json
typo_corrections.schema.json

Best Practices

Confidence scores: Use 0.9+ for high-certainty rules, 0.7-0.9 for moderate, below 0.7 for context-dependent
Context constraints: Always specify context when rules are position-dependent
Examples: Include examples for documentation and testing
Version control: Update metadata.last_updated when modifying rules
Testing: Test rule changes with representative corpus data

Documentation Index

​Rule Files Overview

​File Structure

​Particles (particles.yaml)

​Structure

​POS Tags

​Formality Levels

​Typo Corrections (typo_corrections.yaml)

​Structure

​Error Types

​Context Types

​Morphology (morphology.yaml)

​Structure

​Aspects (aspects.yaml)

​Structure

​Classifiers (classifiers.yaml)

​Structure

​Register (register.yaml)

​Structure

​Negation (negation.yaml)

​Structure

​Homophones (homophones.yaml)

​Structure

​Compounds (compounds.yaml)

​Structure

​Custom Configuration

​Loading Custom Rules

​Extending Rules

​Schema Validation

​Best Practices

​See Also

Rule Files Overview

File Structure

Particles (particles.yaml)

Structure

POS Tags

Formality Levels

Typo Corrections (typo_corrections.yaml)

Structure

Error Types

Context Types

Morphology (morphology.yaml)

Structure

Aspects (aspects.yaml)

Structure

Classifiers (classifiers.yaml)

Structure

Register (register.yaml)

Structure

Negation (negation.yaml)

Structure

Homophones (homophones.yaml)

Structure

Compounds (compounds.yaml)

Structure

Custom Configuration

Loading Custom Rules

Extending Rules

Schema Validation

Best Practices

See Also