# Multilingual Stemmer Stem words with Snowball algorithms in English, German, Spanish, French, Russian, Arabic and more. Browser-local text, stem list, mapping, table or JSON output. ## Run it - **CLI:** `gizza tool multilingual-stemmer "The runners were running quickly. Studies studied studying."` - **Web:** https://gizza.ai/tools/multilingual-stemmer/ - **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/multilingual-stemmer/tool.json ## Inputs - `input` — Text to stem _(field)_ - `language` — Language _(field)_ - `output` — Output format _(field)_ - `min_length` — Minimum word length _(field)_ - `lowercase` — Lowercase before stemming _(field)_ ## Output - Stemmed output (text) ## Query parameters Open the tool pre-filled and auto-run via URL: - `input` — Text to stem - `language` — Language - `output` — Output format - `min_length` — Minimum word length - `lowercase` — Lowercase before stemming Example: `https://gizza.ai/tools/multilingual-stemmer/?input=The%20runners%20were%20running%20quickly.%20Studies%20studied%20studying.&language=english&output=text&min_length=1&lowercase=true` --- ## About this tool **Multilingual Stemmer** turns inflected words into Snowball stems for search, tagging, keyword normalization and small corpus analysis. A stem is the reduced form used as an index key: `running` becomes `run`, Spanish `corriendo` becomes `corr`, and German `Häusern` becomes `haus`. Choose the language that matches the text, then choose how you want the result: - **Stemmed text** preserves punctuation, spacing and line breaks while replacing each word with its stem. - **Unique stems** lists the vocabulary after stemming. - **Form → stem mapping** shows how every distinct surface form was normalized. - **Stem frequency table** counts each stem and the forms that collapsed into it. - **JSON groups + stats** gives machine-readable stem groups for indexing pipelines. Everything runs locally in WebAssembly. No text is uploaded. ### Worked example Input: ```text The runners were running quickly. Studies studied studying. ``` With language **English** and output **Stemmed text**, the result is: ```text the runner were run quick. studi studi studi. ``` For a search index, switch to **JSON groups + stats** to get counts and surface forms per stem. For an analyst checking a vocabulary cleanup, **Form → stem mapping** is usually the most readable view. ### Languages The tool uses Snowball stemming algorithms for Arabic, Danish, Dutch, English, Finnish, French, German, Greek, Hungarian, Italian, Norwegian, Portuguese, Romanian, Russian, Spanish, Swedish, Tamil and Turkish. ### Limits and edge cases - Maximum input: **200,000 characters** per run. - Stemming is language-specific. The wrong language still returns output, but the stems are not meaningful. - Stems are **not lemmas** and may not be dictionary words: `studies → studi` is expected. - The tokenizer treats Unicode letters and digits as words and keeps apostrophes inside contractions. - Use **Minimum word length** to keep short abbreviations such as `AI`, `API` or product codes unchanged. Also available from the gizza CLI and in chat. ## FAQ
Is stemming the same as lemmatization? No. Stemming strips suffixes according to language rules and returns an index key, not a dictionary word. `studies`, `studied` and `studying` all become `studi` in English. A lemmatizer would try to return the dictionary lemma `study`, which usually requires a larger language model or dictionary.
Which language should I choose for mixed-language text? Choose the dominant language, or split the text first and run each language separately. Snowball algorithms are language-specific: German suffix rules applied to Spanish text, or English rules applied to Turkish text, can produce misleading stems even though the tool will still run.
Why did capitalization change? By default, words are lowercased before stemming because Snowball algorithms are defined for lowercase input. This makes `Running` and `running` collapse to the same stem. Turn off **Lowercase before stemming** only when case distinctions are part of the token you need to preserve.
What output should I use for search indexing? Use **JSON groups + stats** if a downstream program needs counts and source forms, or **Unique stems** if you only need the vocabulary. Use **Stemmed text** when you want to feed a normalized text stream into another simple text tool while keeping punctuation and line breaks in place.
Will it handle a full book or a production corpus? It is intended for pasted snippets, keyword lists and small batches, not full-corpus ETL. The per-run cap is 200,000 characters so the browser and chat WebAssembly sandbox stay responsive. For larger corpora, split the text into chunks or run a dedicated indexing pipeline.
## Related tools - [Extract Action Items from Meeting Notes](https://gizza.ai/tools/action-item-extractor/): Extract action items, owners, and decisions from meeting notes or daily notes with deterministic rules. Markdown checklist or JSON, private in-browser. - [Add Line Numbers](https://gizza.ai/tools/add-line-numbers/): Add line numbers to every line of text online, like nl or cat -n — custom start, step, separator, and alignment. Free and private, runs in your browser. - [ANSI Log Renderer](https://gizza.ai/tools/ansi-log-renderer/): Paste ANSI-colored terminal output or CI logs and render them as HTML, or strip escape codes to plain text. Handles 16-color, 256-color, and truecolor SGR codes. - [API response diff](https://gizza.ai/tools/api-response-diff/): Compare two JSON API responses and see only the meaningful changes: ignore request ids, timestamps and UUIDs, match arrays by key or as sets, export a JSON Patch. - [Rust AST Diff](https://gizza.ai/tools/ast-diff/): Compare two Rust source snippets structurally. The tool parses both files, canonicalizes their ASTs, and ignores formatting, whitespace, and comments.