Language Detector: 75 languages, codes and confidence

Find out what language each text is written in. Paste texts, send JSON records with your own IDs, list web pages, or point it at an Apify dataset, and get one result per text with the language name, its ISO codes and how sure the detector is.

What you get for each text

• language: the English name, such as German or Norwegian Bokmål
• code: ISO 639-1 (de) and iso639_3: ISO 639-3 (deu)
• confidence: from 0 to 1
• reliable: true when the confidence reaches your threshold (default 0.5) and leads the next language by at least 0.1. A short text can fall below it: "Bedankt voor je bestelling!" scores 0.40 for Dutch across all 75 languages, and 0.76 when the choice is limited to Dutch, German, Danish and Norwegian Bokmål.
• candidates: the top languages with their confidence, best first (default 3)
• textPreview (the first 120 characters) and charCount
• For web pages, declaredLanguage: the language the page's own html tag states, next to the language of its text

Languages

Afrikaans, Albanian, Arabic, Armenian, Azerbaijani, Basque, Belarusian, Bengali, Bosnian, Bulgarian, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Esperanto, Estonian, Finnish, French, Ganda, Georgian, German, Greek, Gujarati, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Irish, Italian, Japanese, Kazakh, Korean, Latin, Latvian, Lithuanian, Macedonian, Malay, Maori, Marathi, Mongolian, Norwegian Bokmål, Norwegian Nynorsk, Persian, Polish, Portuguese, Punjabi, Romanian, Russian, Serbian, Shona, Slovak, Slovenian, Somali, Southern Sotho, Spanish, Swahili, Swedish, Tagalog, Tamil, Telugu, Thai, Tsonga, Tswana, Turkish, Ukrainian, Urdu, Vietnamese, Welsh, Xhosa, Yoruba and Zulu.

When you know your texts can only be a few languages, list their ISO 639-1 codes. The detector then chooses among those alone, which helps most on short texts.

Input

• Texts: one per entry
• Texts with IDs: JSON objects such as {"id": "review-17", "text": "...", "stars": 5}. Every field except the text is copied to the result, so you can join it back.
• Web pages: the main text of each page is detected. Menus, footers and scripts are left out.
• Apify datasets: each item is one text. The field is found for you (text, markdown, content, body, description, title, review, comment), or name it.
• Limit to languages, candidates per text, reliable at confidence, fewest letters (default 3), fast mode

Output

Dataset: one item per text, in input order, with index and id. A text with fewer letters than the minimum, such as "ok" or "12345", gets language: null and a note. A source that cannot be read gets ok: false and an error, such as HTTP 404 or a record with no text. The other texts still run.

Pricing

Pay per event: one event per text with a detected language. Texts too short to detect and sources that fail are not charged. The run stops when it reaches the maximum charge you set.

How it works

Detection runs inside the actor with the open source Lingua library (Apache 2.0 license). No text is sent to any outside service, and nothing is stored beyond the run's own storage.

Support

Questions and problems: the Issues tab on this page. Replies are written by an AI assistant on behalf of the developer and reviewed by the developer.