Sources & Methodology
Where every part of this collection comes from, which fields are sourced and which are written, and what is still weak. If you only read one section, make it the last one.
Proverbs
| Field | Where it comes from |
|---|---|
| Proverb text and meaning | A Dictionary of English Proverbs and Proverbial Phrases (1870), Project Gutenberg ebook #39281. Public domain. |
| Additional entries and translations | English Wiktionary, licensed CC BY-SA 4.0. |
| French, Spanish and Italian equivalents | Henry G. Bohn, A Polyglot of Foreign Proverbs (1857), Project Gutenberg ebook #51090. Public domain. |
The proverb collection is the best-sourced part of the site: the sayings and their glosses come from printed reference works, not from a model.
One caveat worth stating. Where a proverb carries a translation, that is not always an attested proverb in the other language — some are equivalents that genuinely exist there, others are renderings of the English. The data does not yet distinguish the two, and separating them is work in progress.
Idioms
| Field | Where it comes from |
|---|---|
| The phrase itself | English Wiktionary, used to identify real, attested idioms — plus batches curated by hand. |
| Literal translation, meaning, origin | Written by a language model, then verified. Wiktionary's own gloss is used only as an accuracy signal and is never copied, so no CC BY-SA text enters the database. |
| English idiom origins | Researched against etymonline, World Wide Words, phrases.org.uk and Grammarist. Where the popular story is folklore — get someone's goat, talk turkey — it is labelled as folklore rather than presented as fact. |
This is the part of the site that needs the clearest disclosure. The idiom phrases are attested; their explanations are drafted by a model and then gated by a second pass that rejects anything it cannot support. That process is good enough to be useful and not good enough to be taken on trust.
Other collections
Untranslatable words, false friends, compliments and insults, and tongue twisters follow the idiom pipeline: the words and phrases are real, and the explanations are written and then checked.
How entries are checked
- Attestation. Every idiom is checked against English Wiktionary for whether the phrase exists at all, and whether it is filed under the right language.
- Fabrication and etymology audit. A separate model-as-judge pass looks specifically for invented phrases, wrong language labels, and origins that read as confabulated.
- Hand rewriting. Anything flagged is rewritten or deleted by a person. Roughly two hundred thin or doubtful origins were rewritten this way between August and September 2026.
Corrections we have made
Publishing the error log matters more than publishing the process:
- June 2026 — seventeen idioms that both audits independently judged fabricated were deleted outright.
- June 2026 — one entry (臥虎藏龍) was relabelled from Japanese to Chinese; it was the only CJK entry with no Japanese Wiktionary section.
- June 2026 — nine origins judged likely false — including cloud nine and cold shoulder — were rewritten.
- August–September 2026 — around two hundred thin or unsupported origin notes rewritten across the French, German, Spanish, Italian, Russian, Japanese, Korean, Chinese, Dutch, Swedish, Polish, Portuguese, Arabic, Greek, Hindi and Turkish sets, and several duplicate entries merged.
What is still weak
The parts of this site we would not want you to rely on without checking:
- About sixty-nine idiom origins remain flagged as doubtful by the audit and have not yet been rewritten or removed. They are lower-confidence than the rest and are still on the site.
- Idiom explanations are model-written. Verified, but written — not quoted from a reference work, the way the proverbs are.
- Some proverb translations are renderings, not attested sayings in the target language, and are not yet labelled as such.
- Origins are the weakest field everywhere. Attestation can confirm that a phrase exists; nothing can confirm a story about where it came from.
If you find something wrong, idiomingo.support@gmail.com is the fastest way to get it fixed.