Credits & Data Sources

Spanish Forge is built on open data produced by other people. This page lists those sources, the licences they carry, and what we have added on top of them.

The vocabulary data derived from them — our phrase list, inflection tables and lemma table — is free to download from SpanishForge/phrases-data on GitHub, under the same licences.

1. Vocabulary frequency data

Frequency list derived from hermitdave/FrequencyWords (CC BY-SA 4.0), built from the OpenSubtitles 2018 corpus via OPUS.

• Source: hermitdave/FrequencyWords, the Spanish word list under content/2018/es.
• Licence: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0).
• Upstream corpus: OpenSubtitles 2018, distributed via OPUS.

Because that frequency list is share-alike, the Spanish Forge phrase table — Spanish headword, English gloss, CEFR level, part of speech, gender, frequency rank and regional variants — is itself published under CC BY-SA 4.0, with the attribution above and a licence file alongside the data, in SpanishForge/phrases-data.

The CEFR levels, English definitions, parts of speech, genders and regional variants were assigned by Spanish Forge with the help of large language models. They are our additions and are not part of the upstream frequency list.

2. Lemmatisation data

Matching an inflected word in a text (hablamos, vino, perros) against a headword stored as a lemma (hablar, venir, perro) uses a form-to-lemma table compiled from:

• Source: michmech/lemmatization-lists, file lemmatization-es.txt, by Michal Boleslav Měchura.
• Licence: Open Database License (ODbL) v1.0. The table we compile from it is a derivative database and is published under the same licence in SpanishForge/phrases-data, with a description of every change we made to the source; the surrounding code is not affected.
• Attribution: required by the ODbL and carried in the asset's header comment lines, in its provenance sidecar, and here.

3. Inflection tables

Inflection tables derived from English Wiktionary (CC BY-SA 4.0) via wiktextract / kaikki.org. They are what lets us show a verb's conjugations, a noun's plural and the masculine, feminine and plural forms of an adjective.

• Source: English Wiktionary, Spanish entries, as extracted by wiktextract (Tatu Ylonen) and published by kaikki.org.
• Licence: Wiktionary text is dual-licensed CC BY-SA 4.0 and GFDL (see Wiktionary's copyright page); the table we compile from it is a derivative and is published under CC BY-SA 4.0 in SpanishForge/phrases-data.
• Attribution: carried in the asset's header comment lines, in its provenance sidecar, and here.

Sorting Wiktionary's form labels into the fixed set of slots we show, working out the forms of reflexive verbs such as acostarse from the plain verb's table, and filling the remaining gaps with the help of large language models are our additions. They are not part of the upstream data.

4. What is not covered by these licences

The sentences, stories, audio recordings, images and AI tutor conversations that Spanish Forge produces are original works, and so are the recordings of individual inflected forms. They are not adaptations of the word list, the lemma table or the inflection tables, and they are not covered by share-alike.

5. Mandarin Forge

Spanish Forge is built on the codebase of Mandarin Forge, its sister project for learners of Chinese. The app's structure, the spaced-repetition system and much of the interface come from there; the Spanish content, data pipeline and voices were built for Spanish Forge.

6. Other services

Speech, images and generated text are produced with third-party AI providers. The providers we send data to are listed in our Privacy Policy.

7. Contact

If you have a question about attribution or licensing on this page, or you believe a source is missing from it, write to us at hello@spanishforge.com.