Lexicor

Lexicor is a strictly typed, highly optimized, offline interface for the Open English WordNet.

✨ Features

🚀 Ultra Fast: Microsecond-level lookups (~20µs on disk). WITHOUT ROWID tables, covering indexes, and LRU result caching.
🔒 Strictly Typed: Concept, SpeechPart, RelationType, DomainCategory — no magic strings or integers anywhere.
🧠 Morphology Aware: Automatic stem resolution. "ran" → "run", "better" → "good".
📖 Optional Definitions: Pass withDefinitions: true to get WordNet glosses on every concept — or keep the default lightweight build when you don't need them.
🔗 Rich Relations: Semantic (Concept-to-Concept) and Lexical (Word-to-Word) relations, filterable by type in SQL.
🌲 Recursive Traversal: Walk hypernym/hyponym chains to arbitrary depth via a single SQLite recursive CTE.
⚡ Dual Storage Modes: Disk mode (instant startup) or Memory mode (fastest queries).

📦 Installation

dependencies:
  lexicor: ^0.2.0

🚀 Quick Start

import 'package:lexicor/lexicor.dart';

void main() async {
  final lexicor = await Lexicor.init();           // lightweight, no definitions
  // final lexicor = await Lexicor.init(withDefinitions: true); // full mode

  final result = lexicor.lookup('bank');
  print('Found ${result.concepts.length} concepts for "${result.query}"');

  for (final concept in result.concepts) {
    print('[${concept.part.label}] ${concept.category.label}');
    if (concept.definition != null) print('  ${concept.definition}');

    final hypernyms = lexicor.related(concept, type: RelationType.hypernym);
    for (final rel in hypernyms.items) {
      print('  -> is a type of: ${rel.word}');
    }
  }

  lexicor.close();
}

📖 Usage Guide

Initialization

// Light mode — no definitions (~28 MB asset, default)
final lexicor = await Lexicor.init();

// Full mode — definitions + usage examples (~40 MB asset)
final lexicor = await Lexicor.init(withDefinitions: true);

// Memory mode — copy DB into RAM for fastest repeated queries
final lexicor = await Lexicor.init(
  mode: StorageMode.inMemory,
  withDefinitions: true,
);

Flag	Asset	`Concept.definition`	`examples()`	Best for
`withDefinitions: false` (default)	`dictionary.sqlite` ~28 MB	`null`	`[]`	Mobile, CLI, embedding
`withDefinitions: true`	`dictionary_full.sqlite` ~40 MB	populated	sentences	Dictionaries, NLP

Definitions

final lexicor = await Lexicor.init(withDefinitions: true);
final result = lexicor.lookup('bank');
for (final concept in result.concepts) {
  print(concept.definition);
  // "a financial institution that accepts deposits..."
  // "a long ridge or pile..."
}

concept.definition is always null in light mode — safe to use without checking the mode:

final def = concept.definition ?? '(no definition)';

Usage Examples

final lexicor = await Lexicor.init(withDefinitions: true);
final concept = lexicor.lookup('run').primary!;

for (final example in lexicor.examples(concept)) {
  print(example); // "she ran to catch the bus"
}

Returns an empty list in light mode.

Synonyms

final concept = lexicor.lookup('car').primary!;
final syns = lexicor.synonyms(concept);
// ['car', 'auto', 'automobile', 'motorcar']

These are strict synonyms — words that share the exact same synset. For broader similarity, use related(concept, type: RelationType.similar).

Recursive Traversal

Walk a relation chain to arbitrary depth in a single DB call:

final dog = lexicor.lookup('dog').primary!;
final chain = lexicor.traverse(dog, RelationType.hypernym, maxDepth: 5);

for (final word in chain) {
  print('${'  ' * word.depth}${word.word}');
}
// canine
//   carnivore
//     placental
//       mammal
//         ...

traverse uses a SQLite recursive CTE — the entire hierarchy walk is done in one round-trip. Only works for relations where RelationType.isRecursive is true (hypernym, hyponym, partMeronym, etc.).

Relationships

final rels = lexicor.related(concept);

// SQL-level type filter — no wasted data transfer
final hypernyms = lexicor.related(concept, type: RelationType.hypernym);

// Or filter the result object
final semantic = rels.semantic;            // Concept-to-Concept links
final lexical  = rels.lexical;             // Word-to-Word links
final antonyms = rels.byType(RelationType.antonym);

Batch Lookup

final results = lexicor.lookupBatch(['run', 'walk', 'swim', 'run']);
// 'run' is only queried once; the second entry is served from cache.
results['run']?.concepts.length;

Morphology

// Automatic during lookup:
lexicor.lookup('running').resolvedForms; // ['running', 'run']

// Direct lemmatization:
lexicor.lemmatize('ran', SpeechPart.verb);        // 'run'
lexicor.lemmatize('better', SpeechPart.adjective); // 'good'

Flutter Integration

import 'package:path_provider/path_provider.dart';
import 'package:flutter/services.dart';
import 'package:lexicor/lexicor.dart';

Future<Lexicor> initLexicor({bool withDefinitions = false}) async {
  final assetName = withDefinitions ? 'dictionary_full.sqlite' : 'dictionary.sqlite';
  final dir  = await getApplicationDocumentsDirectory();
  final file = File('${dir.path}/$assetName');

  if (!file.existsSync()) {
    final data = await rootBundle.load('packages/lexicor/assets/$assetName');
    await file.writeAsBytes(data.buffer.asUint8List());
  }

  return Lexicor.init(customPath: file.path, withDefinitions: withDefinitions);
}

🛠 Building the Database

Place your oewn.sqlite (~165 MB source file) at tool/db/oewn.sqlite, then:

# Light build only (~28 MB)
dart run tool/build_database.dart tool/db/oewn.sqlite

# Full build only (~40 MB)
dart run tool/build_database.dart tool/db/oewn.sqlite --full

# Both at once
dart run tool/build_database.dart tool/db/oewn.sqlite --all

# Or via melos
melos run build
melos run build:full
melos run build:all

📊 Database Stats

Component	Light	Full	Description
Words	3.0 MB	3.0 MB	~150k unique lemmas
Concepts	1.36 MB	1.36 MB	~120k Synsets
Senses	3.4 MB	3.4 MB	~210k Word-Concept pairs
Semantic Relations	3.7 MB	3.7 MB	Hypernyms, Holonyms, Entailments
Lexical Relations	4.0 MB	4.0 MB	Antonyms, Derivations
Indexes	~11 MB	~11 MB	`COLLATE NOCASE` indexes
Definitions	—	~10 MB	Synset glosses (~120k rows)
Examples	—	~2 MB	Usage sentences from `samples` table
Total	~28 MB	~40 MB

📂 License

Package License: MIT
Database: Open English WordNet 2025 (v2.3.2), licensed CC BY 4.0.

Commercial use requires attribution to the Open English WordNet project in your app's About / Licenses section.