What Language Is This?

How the detector recognises a language

The first clue is the alphabet. Some writing systems belong to one language, or nearly so: Hangul is Korean, the Thai script is Thai, Greek letters are Greek, and kana mixed with Chinese characters is Japanese. When the text is in one of those scripts, the answer is almost certain.

Latin, Cyrillic, Arabic and Devanagari letters are shared by dozens of languages, so the detector looks at which letter sequences turn up most often. Every language has its own habits. English is full of "the", "ing" and "and"; Polish of "prz", "ów" and "dz"; German of "sch" and "ein"; Spanish of "que" and "ción". The text is compared with the typical sequences of about 80 languages, and the closest match wins.

When the answer is less certain

The confidence shown with each answer says how far ahead the winner was. Three situations bring it down:

• Very short text. A greeting or a name does not hold enough letter patterns to decide. Two or three full sentences are usually plenty.
• Close relatives. Bosnian, Croatian and Serbian in Latin letters are written almost the same way, as are Indonesian and Malay. Hindi, Marathi and Nepali share the Devanagari script and many words. When two such languages run close, both are named.
• Mixed text. An English email that quotes a French contract is read as a whole, and the language with more text wins. Paste the part you are unsure about on its own.

The detector names languages, not regional varieties. It cannot tell Brazilian from European Portuguese, or the Spanish of Mexico from that of Spain; their spelling differs too little.

Checking a document, not just a sentence

Drop a PDF, Word, Excel, PowerPoint, InDesign (IDML) or EPUB file, or a plain text file, and the detector reads its text in your browser and names the language from the document as a whole. It is a quick check before you send a file for translation, and a way to catch an attachment that is not in the language you expected.

A scanned document has no text to read. For a scan or a photo, first get the text out with Image to Text, then paste it here.

From knowing to translating

When you upload a document to DocTranslator, you choose the language it is written in and the one you want, from 108 languages. Once you know the first, the rest is a few clicks, and the translated file keeps the layout of the original.