How the language detector works
The Language Detector analyzes one text input in your browser. It returns a Primary Language, up to eight Ranked Candidates, normalized heuristic scores, and warnings when the text is short or ambiguous.
It checks Unicode scripts for languages such as Russian, Greek, Arabic, Hebrew, Hindi, Chinese, Japanese, Korean, and Thai. For Latin-script languages, it compares word matches against curated stopword-frequency profiles. The resulting values are normalized and ranked.
Enter text or upload a file
- Type or paste text into Paste text to detect its language.... Analysis updates automatically as you edit the field.
- To use a file, select Upload and choose a
.txt,.csv,.log, or.mdfile. - You can also drag a file onto the text input and release it. The file is read locally in your browser.
Read the results and statistics
After text is entered, the page shows the detected Primary Language, its displayed confidence value, and any reliability warnings. The statistics line reports the number of analyzed words and scanned letter characters.
- Ranked Candidates lists up to eight candidates in score order.
- Supported Languages shows the language signals covered by the detector.
- Copy copies the displayed ranked candidates. CSV downloads them as
language-detection-report.csv. Both do nothing when there are no candidates.
When no language is identified
Read the warnings before relying on a result:
- Fewer than 5 words can trigger a very short-text warning. Fewer than 15 words can trigger a short-text reliability warning.
- Close scores for the top two candidates can produce an ambiguity warning.
- The Primary Language panel can show No language could be confidently identified from this text.
- The table can show No matching language signal found. Check Supported Languages when this happens. Languages outside that list are not covered.
Limitations and browser-local processing
Latin-script detection uses curated, non-exhaustive stopword lists. This is a fast client-side heuristic, not a full statistical language model. Treat the displayed scores as normalized heuristic values, not definitive accuracy measurements.
Detection and local .txt reading run in the browser. The page code does not upload the entered text to a server.
Frequently Asked Questions (FAQs)
How does the Language Detector work?
It combines Unicode script checks with curated stopword-frequency profiles, then ranks up to eight candidates by normalized heuristic score.
Why does short text give unreliable results?
A handful of words rarely contains enough distinctive vocabulary to separate similar languages confidently. Text under 15 words is flagged as short and less reliable; under 5 words is flagged as very unreliable.
Which languages are supported?
See the Supported Languages panel above for the complete, explicit list. Any language not listed there is outside this tool's coverage and cannot be reliably identified as its own language.
Is my text sent to a server?
Detection and local text-file reading run in your browser. The page code does not upload the entered text to a server.