cstr/de-wiktionary-semantic

German Wiktionary - Normalized SQLite Database with semantic web data etc

This is a lossless, fully normalized SQLite database of German Wiktionary, capturing every field from the cstr/de-wiktionary-extracted dataset.

🎯 Key Features

• ✅ 100% Lossless: ALL fields captured including:
  • 🔗 Wikilinks in definitions (semantic connections)
  • 📝 Qualifiers (e.g., "archaic", "Swiss", "informal")
  • 🏷️ Sense IDs (unique identifiers)
  • 🌐 Wikidata IDs (for semantic web linking)
  • 📚 Attestations (historical citations)
  • 🎭 Head templates (morphological data like Genus/Plural)
  • 📖 Info templates (structured metadata)
• ⚡ Fast Queries: Fully indexed schema for sub-20ms queries
• 🔗 Complete Semantic Web: All relations preserved with sense-level granularity
• 📱 Mobile-ready: Optimized for sqflite (Flutter) and local DB use cases

📊 Database Statistics

• Entries: 970,801
• Word Senses: 3,098,364
• Definitions (Glosses): 3,087,300
• Wikilinks: 0
• Sense IDs: 3,098,364
• Qualifiers: Embedded in senses
• Translations: 1,131,251
• Word Forms: 6,100,090
• Head Templates: 0
• Pronunciations: 2,327,762
• Examples: 427,322
• Attestations: 0
• Wikidata IDs: 0
• Synonyms: 161,563
• Antonyms: 76,054
• Hypernyms: 133,059
• Hyponyms: 0

🏗️ Database Schema (40+ Tables)

New Tables (vs Previous Versions)

• head_templates: Morphological templates (Crucial for German inflection/gender)
• entry_wikipedia: Wikipedia cross-references
• sense_links: Wikilinks in definitions
• sense_raw_tags: Unstructured tags
• sense_wikidata: Wikidata identifiers
• sense_wikipedia: Wikipedia at sense level
• attestations: Historical citations
• info_templates: Structured metadata

Core Tables

• entries: Core word data with etymology
• senses: Definitions with qualifier, senseid, head_nr
• translations: Multi-language translations
• examples: Usage examples
• semantic relations: hypernyms/hyponyms/meronyms/holonyms/coordinate_terms

📖 Usage

Download

from huggingface_hub import hf_hub_download
import sqlite3
import gzip
import shutil

# Download compressed database
db_gz_path = hf_hub_download(
    repo_id="cstr/de-wiktionary-semantic",
    filename="de_wiktionary_normalized_full.db.gz",
    repo_type="dataset"
)

# Decompress
db_path = db_gz_path.replace('.gz', '')
with gzip.open(db_gz_path, 'rb') as f_in:
    with open(db_path, 'wb') as f_out:
        shutil.copyfileobj(f_in, f_out)

# Connect
conn = sqlite3.connect(db_path)

Example Queries

# Get definition with wikilinks for "Hund"
cursor.execute('''
    SELECT g.gloss_text, GROUP_CONCAT(l.link_text, ', ') as links
    FROM entries e
    JOIN senses s ON e.id = s.entry_id
    JOIN glosses g ON s.id = g.sense_id
    LEFT JOIN sense_links l ON s.id = l.sense_id
    WHERE e.word = ? AND e.lang = 'Deutsch'
    GROUP BY g.id
''', ('Hund',))

# Find Wikidata ID for a sense
cursor.execute('''
    SELECT e.word, w.wikidata_id
    FROM entries e
    JOIN senses s ON e.id = s.entry_id
    JOIN sense_wikidata w ON s.id = w.sense_id
    WHERE e.word = ?
''', ('Katze',))

📜 License

CC-BY-SA 4.0 (same as source)