Loading page
About this tool
Segment Chinese text, split English and Hindi, detect mixed language, and export live NLP-style token results in the browser.
Chinese is written without spaces, so before you can count words or process a sentence you have to decide where each word ends — this tool does that using the browser's built-in Intl.Segmenter with word granularity and the zh-Hans locale. It first splits your text into runs by script, detecting Chinese, Hindi, Latin and numeric characters, then segments each run with the right locale, so a mixed sentence is tokenised correctly rather than being forced through one rule. Each token is listed with its script, character length and an optional gloss you supply, and the whole result exports as TXT or JSON.
Open Chinese Word Segmentation & Multi-Language Split Tool on AltFTool — it loads instantly in your browser.
Type or paste into the Input Panel — it takes Chinese, English, Hindi or mixed-language text, and there is no file upload. Anything past 12000 characters is trimmed, with the note Input limited to 12000 characters for responsive parsing; otherwise the panel reads Updates as you type. No backend required.
Nothing needs to be run. Language Detection names the script and counts Chinese, English, Hindi and Number characters, Comparison View sets Original Text against the Segmented Version whose tokens are joined by a pipe, and Segmentation Output lists each token with its script label — narrow that list with Search tokens, scripts, or meanings, or add your own glosses under Meaning Layer as one mapping per line (中文 = Chinese language).
Copy puts the segmented text on the clipboard, TXT and JSON save language-segmentation-result.txt and language-segmentation-result.json, and Save files a snapshot under History, which keeps the last 12 and reloads one when you click it. Clear empties the input; Clear history empties the list.
Text is split into runs of Chinese, Hindi, Latin, numeric and punctuation characters first, so each run is tokenised with its own locale rules.
Intl.Segmenter is the browser's built-in Unicode segmentation, the same engine used for text selection, with a regex word-match fallback where it is unavailable.
Paste a simple list of term = meaning pairs and every matching token carries its meaning through into the token table and the JSON export.
Because written Chinese has no spaces between words, and a word can be one, two or more characters long. Without segmentation, a program cannot tell whether two adjacent characters form a single word or two separate ones, which breaks word counts, search indexing and translation memory.
Chinese with the zh-Hans locale, Hindi with hi-IN, and English or other Latin-script text with en-US, plus separate handling for numbers and punctuation. When more than one of these appears it is reported as mixed language, and each run is still segmented under its own rules.
Yes, input is capped at 12,000 characters, and anything beyond that is trimmed with a truncation notice rather than silently dropped. Recent analyses are kept in a local history of up to 12 entries.
It matches the segmentation the browser itself uses, which handles ordinary running text well but has no domain dictionary, so proper nouns, brand names and technical terms may be split apart. Dictionary-based segmenters can do better on those cases; this one has the advantage of needing no model download and running on your text locally.