Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

Processed Wikipedia SQLite Databases for Wikilite

This dataset provides pre-processed SQLite databases of Wikipedia articles for use with the Wikilite tool. These databases allow you to quickly and efficiently search and access Wikipedia content offline using Wikilite's lexical and semantic search capabilities.

Supported Languages

Currently, the dataset includes databases for the following languages:

  • Sardinian (sc)
  • Italian (it)
  • Spanish (es)
  • English (en)
  • German (de)
  • Chinese (zh)
  • Arabic (ar)

More languages may be added in the future.

Dataset Structure

Each language is stored as a separate compressed file (.db.gz). There are two types of databases available:

  1. Lexical Only: Located in the lexical/ directory. These support standard FTS5 full-text search (exact word matching) and are lighter in size.
  2. Semantic & Lexical: Located in the root or specific model directories. These databases contain embedded GGUF models directly within the file structure, enabling both keyword search and vector-based conceptual search without external dependencies.

How to Use This Dataset

1. Command Line & Web Interface (Linux, macOS, Windows, Termux)

Wikilite runs primarily as a command-line tool that can also serve a local web interface.

  1. Install Wikilite
    Download the precompiled binary for your operating system from the Wikilite Releases page and extract it.

  2. Download a Database
    You can download the .db.gz files manually from this repository and extract them, or use Wikilite's built-in tools to handle it for you:

    • Interactive Wizard: Run ./wikilite without arguments to launch a wizard that guides you through selecting and downloading a database.
    • Setup Command: Run ./wikilite --setup to view and download available databases automatically.
  3. Run Wikilite
    Once you have a database file (e.g., wikilite.db), you can use it in three ways:

    • Interactive Mode:

      ./wikilite
      

      (Follow the on-screen prompts)

    • Command Line Search:

      ./wikilite --cli --db <file.db>
      
    • Web Interface:

      ./wikilite --web --db <file.db>
      

      Access the interface at http://localhost:35248 in your browser.

2. Android Application

A native Android application is available that uses these exact databases.

  1. Download the Android App APK.
  2. External Storage: If you manually download a wikilite.db file from this repository and place it on your external SD card, the app will detect and use it automatically.
  3. In-App Download: If no database is found on launch, the app provides an option to download these pre-built databases directly.

About Wikilite

Wikilite is a self-contained tool for creating and browsing local SQLite databases of Wikipedia articles.

Features

  • Lexical Search: Uses FTS5 for efficient, exact keyword matching.
  • Semantic Search: Uses embedded GGUF models (via llama.cpp integration) contained within the database files to find content based on meaning, synonyms, and conceptual similarity.
  • Offline Operation: Complete functionality without internet connectivity.
  • Cross-Platform: Available for Linux, macOS, Windows, Termux, and Android.

Semantic Search Capabilities

The semantic search functionality employs text embeddings to handle:

  • Query misspellings and typographical errors.
  • Conceptual similarity despite different terminology.
  • Synonym and related term matching.
  • Morphological variations (plurals, verb tenses).

Note: The semantic databases in this repository are fully self-contained. No external model files or API keys are required.

Contributing

If you would like to contribute databases for additional languages, please feel free to submit a pull request.

Acknowledgments

  • Wikipedia: For providing the valuable data.
  • SQLite: For the robust database engine.
  • LLaMA.cpp: For enabling the internal generation of embeddings.
  • Wikilite: For the software powering these datasets.
Downloads last month
1,430