About Phrasary

What it is

Phrasary is a reference site. Each language has one page: its common phrases, its writing system where one is recorded, and its core vocabulary. It is meant to be consulted rather than read — you want to say thank you in Khasi and need the word and how to say it.

Every phrase carries a pronunciation respelling with the stressed syllable in capitals: khu-BLEI, KUM-no, na-MAS-te. That notation is the point of the site. It is the part the large translation tools do not give you, and for many of these languages the part no one else typesets at all.

Which languages, and why

There are 49 languages at launch. They were chosen by demand — the languages people search for and then find poorly served — with major languages excluded, since those are covered well elsewhere. A language is included only when its data clears a floor of forty entries, so no page is a stub.

The set grows. Languages are added in the same way, and a language whose data does not yet pass the checks below is held back rather than published thin.

How pages are checked

Every page passes mechanical checks before the site can be built. Some guard against thin or repetitive pages: each page must carry at least three hundred words and be at least forty percent unique against the rest of the site. Others guard the data itself: no page may be mostly English repeated back as if it were a translation; no row may be written in a script other than its page's own; and no Spanish, French or Portuguese weekday, number or greeting may appear in a language that has no history of contact with those languages — the signature of a row copied in from the wrong dataset. A language that fails any of these is held back until the data is fixed.

Those checks catch what a machine can catch. They do not judge whether a translation is right; only a speaker can. Where an entry is wrong, it is wrong in a way a speaker will recognise, and corrections are how the data improves — how to send one.

Editorial notes

Some pages carry a short editorial note about the language. A note is published only after a speaker has verified it, and every note cites its sources on the page. Until it is verified, it does not appear — research in progress is not published as fact.

The translate box

The translate box on each page is machine translation from Google. It is labelled as such and kept apart from the checked entries. When what you type is already in the page's phrasebook, that checked entry is shown above the machine's answer.

What it does not claim

Phrasary does not state speaker counts, where a language is spoken, or ratings of any kind, because its data does not carry them and the site will not invent them. Related languages are drawn from a family, branch and region classification we maintain following standard references, and every relationship shown says which of those it is.

No flags. Several of the languages here have none, and languages are not countries.