DATA & LICENSES

Open-Source Attributions

1. GCIDE Dictionary

GCIDE is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License v3 or later.

GCIDE is distributed WITHOUT ANY WARRANTY. Official home: https://gcide.gnu.org.ua/

2. Kaikki.org & Wiktionary

Definitions and examples from https://kaikki.org/ – a machine‑readable dictionary derived from Wiktionary.

Wiktionary: Base linguistic data sourced via Wiktionary under the CC BY-SA 3.0.

Based on data extracted on 2026-04-25 from the enwiktionary dump dated 2026-04-01 using wiktextract.

3. ConceptNet & Open English WordNet

ConceptNet: Data from ConceptNet 5, licensed under CC BY-SA 4.0 from http://conceptnet.io.

Open English WordNet: Synonyms and word relations based on Princeton WordNet 3.0.

4. Translation & Additional Sources

Google ML Kit On-Device Translation: Credit: Google LLC. Licensed under the Google ML Kit Terms of Service.

CC-CEDICT: Chinese dictionary data via MDBG. Licensed under CC BY-SA.

CMU Pronouncing Dictionary: Phonetic transcriptions. Public domain.

Tatoeba & Moby Thesaurus: Sentences from Tatoeba.org under CC BY 2.0 FR. Moby synonyms (Public domain).

5. AI Language Models

Meta Llama 3: Licensed under the Meta Llama 3 Community License. AI enrichment provided by Meta.

OpenAI: API corrections and linguistic refinement.

Ollama: Licensed under the MIT License.

6. SQLite Database

Your app runs on SQLite. It is in the public domain.

Combined Database: Derivative work distributed under the GNU GPL v3.0.

7. NLP & Utility Libraries

spaCy (MIT) & NLTK (Apache 2.0): Natural Language Processing.

PyEnchant: Spell checking. LangDetect: Language detection.

Requests: HTTP library for API communication.

8. AI Attributions & Privacy

Meta Llama 3.2: Used for metadata generation (POS, Synonyms, Antonyms) and definition repair. https://llama.meta.com — Llama 3.2 Community License.

Ollama: Local inference engine used to process the dictionary database. https://ollama.com

AI Data Processing: Enhanced using Meta Llama 3.2 via Ollama. Privacy Guard: All processing was performed locally and offline.

Academic Standards: Vocabulary frequency and grading filters are cross-referenced with standard CEFR benchmarks, including the Oxford 5000. Linguistic frequency sorting powered by the Zipf-scale distribution model.

9. Additional AI Models

Microsoft Phi-3: Developed by Microsoft. Licensed under the MIT License.

This model contributes to on-device linguistic refinement and academic quality assurance.

10. Frontend & UI Libraries

Fonts: Inter (OFL), Space Grotesk (OFL), and Gentium Plus (OFL). See https://scripts.sil.org/OFL.

Icons: Icons by Lucide (https://lucide.dev) – ISC License.

Components: Radix UI (MIT License) and other React components used are licensed under the MIT License.

1 / 10

GNU GENERAL PUBLIC LICENSE

Version 3, 29 June 2007

Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.