LangDex

How LangDex Works

A graph-based architecture that connects words, meanings, and languages. From language families down to individual word forms.

The Data Layers

From Languages to Words

Six core layers form the foundation of LangDex's linguistic graph. Each serves a specific purpose in connecting the world's languages.

Languoids

The foundation layer. 27,000 language families, languages, and dialects organized in a tree structure from Glottolog. Every language has localized names in 200+ locales.

27K languoids • 5.4M localized names

Concepts

Cross-lingual meaning anchors from PanLex. A concept like 'eat' connects words across 5,700+ languages. The semantic hub that enables universal translation.

2M+ concepts • 5,700+ languages

Lexemes

Core dictionary entries. Each lexeme includes the lemma, language, part of speech, pronunciation (IPA), audio, etymology, frequency rank, and morphological forms.

15M+ lexemes • Multiple sources merged

Senses

Fine-grained definitions for each lexeme. Handles polysemy with domain tags (medical, legal), register info (formal, slang), and example sentences.

30M+ senses • Domain & register tagged

Characters

CJK logographic characters with stroke counts, radicals, readings, and meanings. Kanji, Hanzi, and Hanja with JLPT/HSK level tags.

100K+ characters • 500K+ readings

Word Index

Fast form→lexeme lookup table. Search any inflected form and instantly find the base lexeme. Powers sub-millisecond dictionary lookups.

100M+ word forms indexed

Data Flow

How a Query Traverses the Graph

When you search for a word, LangDex traverses multiple layers to return rich, connected results.

Search Input

User enters a word in any language or script. The word index identifies matching lexemes in milliseconds.

Lexeme Lookup

Retrieve the full lexeme with pronunciation, etymology, frequency, and all morphological forms.

Sense Expansion

Fetch all definitions with domain/register context. Link to example sentences from Tatoeba.

Concept Mapping

Map to cross-lingual concepts. Find translations in any of 5,700+ languages via PanLex IDs.

Relation Traversal

Explore synonyms, antonyms, cognates, etymological ancestors, and derived forms.

Enrichment

Add character breakdowns for CJK, phoneme data, idioms, and WordNet semantic relations.

Data Sources

Built on Open Linguistic Data

LangDex aggregates and structures the world's best open linguistic datasets into a unified graph.

Core Dictionaries

Wiktionary (Kaikki)

51GB of structured entries from 20 language editions

JMDict

214K Japanese-English entries with readings and examples

CC-CEDICT

Chinese-English dictionary with pinyin and definitions

Cross-Lingual

PanLex

11GB of translation pairs across 5,700+ languages

Glottolog

27K languoids in a complete genealogical tree

CLDR

Localized language names in 200+ locales

Enrichment

UniMorph

Morphological forms for 187 languages

IPA-Dict

4.7M pronunciation transcriptions

PHOIBLE

Phoneme inventories for 3K language varieties

Examples & Usage

Tatoeba

10M+ parallel sentences in 400+ languages

OpenSubtitles

Word frequency data from 190 languages

WordNet

120K synsets with semantic relations

API Access

Query the Graph

Access LangDex's data through a simple, powerful API. Three lenses filter results for your use case.

api.langdex.io
GET /v1/lexemes/search?q=taberu&lang=ja&lens=learn

{
  "lexeme": {
    "id": "ja-taberu-verb",
    "lemma": "食べる",
    "reading": "たべる",
    "language": "ja",
    "pos": "verb",
    "ipa": ["/tabeɾɯ/"],
    "frequency_rank": 342,
    "jlpt_level": "N5"
  },
  "senses": [
    {
      "definition": "to eat",
      "domain": ["general"],
      "examples": ["朝ご飯を食べる"]
    }
  ],
  "concepts": ["panlex:eat-12345"],
  "translations": {
    "en": ["eat", "consume"],
    "de": ["essen"],
    "fr": ["manger"]
  }
}

Ready to Explore?

Join our waitlist to get early access to LangDex's API and start building.