Attribution

Where the dictionary comes from, and what it costs to use

NoaLingua's dictionary, grammar tables, examples and frequency lists all come from open datasets, downloaded to your machine. Every one is listed below with its licence. One obvious candidate is deliberately not bundled, because its licence forbids commercial use — and that rule is enforced in the code, not just intended.

Datasets used by NoaLingua, with licences
Source	Licence	What it provides
Tatoeba	`CC BY`	Real bilingual example sentences — the ones shown on every word card. Downloaded as a corpus per language pair, about 8 MB, and searched locally rather than over an API.
Wiktionary, via kaikki.org	`CC BY-SA`	Definitions, IPA pronunciation, register labels and word forms. The largest download by far. English is 475 MB, Chinese 145 MB, French 54 MB. Serbian and Croatian have no extract at all.
UniMorph	`Permissive`	Complete conjugation and declension paradigms. Present for 24 of the 28 languages. French and Italian contain verbs only; Serbian and Croatian share one Serbo-Croatian dataset.
FreeDict / WikDict	`Open`	The offline bilingual dictionary — what answers a hover. Per language pair, and small enough to download almost instantly.
CEFR-J	`Free for commercial use with citation`	Level bands used by the guided A1–C1 course. Cited as the licence requires.
FrequencyWords (Hermit Dave), from OpenSubtitles 2018	`MIT`	The bundled top-1500 frequency lists that drive uncommon-word highlighting and the guided course. Bundled inside the extension for 7 languages, never fetched over the network.
punct-pcs-47lang (anchor-flux)	`Apache-2.0`	Restores full stops and commas in auto-generated captions that have none. 56 MB covering 47 languages in one file, so it downloads once per profile rather than per language.

The one that is missing

Why the Kelly lists are not in here

The Kelly word-frequency lists are the obvious thing to bundle for a “most common words” course. They are licensed CC BY-NC-SA — non-commercial — and NoaLingua charges money.

That makes bundling them a licence violation, so they are not bundled. The rule the project holds itself to is stated plainly: only material whose licence permits commercial redistribution ships in the package. Everything else may be linked to, never bundled.

The frequency lists that are shipped come from Hermit Dave’s FrequencyWords, derived from OpenSubtitles and MIT-licensed — which is why uncommon-word highlighting exists for seven languages rather than for every language Kelly covers. That is a real feature cost, accepted knowingly.

The rule is enforced in three separate modules of the extension rather than being a note in a document, and the whole dataset list was re-audited when the product started charging. Nothing had to be removed, because the rule was held while it was still cheap to hold.

Why this page exists at all Partly because CC BY and CC BY-SA require attribution and this is it. But mostly because “where does your dictionary data come from” is a question a serious learner should ask of any tool like this, and most cannot answer it.

What this means for you

The data is not ours, and we do not pretend it is. NoaLingua is the machinery that fetches, indexes and presents it — the linguistic content is the work of the projects above.
Gaps in coverage are usually gaps upstream. Serbian has no offline dictionary because no Wiktionary extract exists for it, not because we chose to skip it. The coverage matrix shows which is which.
Everything runs on your machine after download. These are files, not services — which is why the product keeps working offline and why the packs are as large as they are.

Try it on the next video you were going to watch anyway.

Free, with every feature. No account, nothing to cancel, and your deck stays on your machine.

Add to Chrome — Free Add to Chrome The Chrome Web Store still lists this as LinguaCue — same extension, rename in progress.

What happens after you install Pricing What it cannot do