Detect the language of a text

Identifying a language from text has been a solved problem for decades and the solution is surprisingly simple: every language leaves a statistical fingerprint in the three-letter sequences it uses. Spanish is full of que, ent, ada; English of the, ing, and; German of sch, ich, der. Comparing how often those sequences appear in your text against each language gets it right with very little material.

That comparison is light enough to run as you type. One full sentence is usually enough; below twenty characters the result stops being reliable, and with isolated proper nouns or place names no method works.

Languages with their own script (Greek, Russian, Japanese, Korean, Arabic, Hebrew, Thai) are identified from the character range rather than the sequences, and there accuracy is essentially perfect.