How to use the language detector
Enter a passage to see its primary language and a ranked list of up to eight candidates. The Language Detector combines Unicode script checks with curated stopword sets, then displays normalized confidence scores and warnings for short text or close scores.
Results update as you edit Text to analyze. The page tool code does not send the entered text to a server.
- Enter or paste text into Text to analyze. The results change with the field.
- Select Upload to choose a
.txt,.csv,.log, or.mdfile. - Check Primary Language, the word and letter counts, and any warning above the results.
- Select Copy for the displayed candidates, or select CSV to download
language-detection-report.csv.
How to identify the language of text
The detector checks Cyrillic, Greek, Arabic, Hebrew, Devanagari, Han, Hangul, Hiragana, Katakana, and Thai characters. For Latin-script text, it tokenizes the words and counts matches against a stopword set for each supported language. It combines those counts, normalizes the scores, and sorts the candidates.
Input: Bonjour tout le monde
Primary Language: French
Confidence: 100.0%When language detection is uncertain
Confidence is a normalized heuristic score, not a measured accuracy percentage. Text with fewer than 5 words gets a very short warning. Text with fewer than 15 words gets a short-text warning. The tool also reports ambiguity when the top two candidates are less than 10 percentage points apart.
- Ranked Candidates contains at most eight entries.
- Text with no matching signal can show "No matching language signal found."
- Languages outside the Supported Languages panel are outside the detector's coverage.
- For Latin-script text, words absent from the curated lists do not add a language signal.
Where language detection comes up
A support editor can paste a customer message into Text to analyze and check the primary result before routing the message to a language queue.
A translator can inspect an excerpt before opening the matching project or reference material. The short-text warning marks excerpts that contain fewer than 15 words.
A developer can upload a text or log file, review the candidate ranking, and select CSV to save the table as language-detection-report.csv.
Choose the right text analysis tool
Use the Text Analysis Tools category when you need a different kind of text check. Choose Homoglyph Detector for Cyrillic, Greek, or Armenian characters that resemble Latin letters. Choose Invisible Character Detector for zero-width spaces, control codes, or bidi overrides. Choose Text Similarity Checker when you need to compare two texts by word overlap.
Frequently asked questions
How does the Language Detector identify a language?
It checks Unicode scripts and curated stopword matches, then ranks the resulting language signals by normalized heuristic score.
How many words does language detection need?
The warning thresholds use word count. Fewer than 5 words triggers a very short warning, and fewer than 15 words triggers a short-text warning.
What languages can this language detector recognize?
It recognizes the languages listed in Supported Languages, including Latin-script languages plus Russian, Greek, Arabic, Hebrew, Hindi, Chinese, Japanese, Korean, and Thai.
Does the Language Detector upload my text?
No. Detection runs in the browser, and this page's tool code does not upload the entered text to a server.