Grammar and Spelling Checker: bulk texts and web pages

Proofread many texts at once. Paste texts, send JSON records with your own IDs, list web pages, or point it at an Apify dataset, and get every spelling, grammar, punctuation and style issue with suggested fixes, counts by kind, and a corrected version of each text.

Checking is done by LanguageTool, the open source proofreading engine (LGPL 2.1 or later), running inside the actor. Your text is never sent to an outside service.

What you get for each text

• issueCount, issuesPer100Words and issuesByKind (spelling, grammar, punctuation, typography, style, casing, confused words, meaning, other)
• issues: each with message, kind, category, ruleId and ruleDescription, offset and length (where it sits in your text), problemText, context (the words around it) and up to 5 replacements
• correctedText: the text with the first suggestion of each issue applied, and correctionsApplied
• language and languageCode used, wordCount
• Every other field of a record you sent, such as your id

For example, "This sentence have a error. Their going to the park tomorow, and they will bring there dog." returns 4 issues and the corrected text "This sentence has an error. Their going to the park tomorrow, and they will bring their dog."

Languages

Arabic, Asturian, Belarusian, Breton, Catalan, Chinese, Crimean Tatar, Danish, Dutch, English, Esperanto, French, Galician, German, Greek, Irish, Italian, Japanese, Khmer, Persian, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, Tagalog, Tamil and Ukrainian, 31 in all, with regional variants such as en-US, en-GB, de-DE, de-AT, de-CH, pt-PT and pt-BR. Rule coverage differs by language.

Leave the language on auto to detect it per text, or give a code. A regional code also checks regional spelling: colour and grey are flagged with en-US and accepted with en-GB.

Input

• Texts, texts with IDs ({"id": "...", "text": "..."}), web pages (main text, with menus and footers left out), Apify datasets (text field found for you, or named)
• Language (auto by default), preferred variants for auto (such as en-GB), the writer's first language (turns on false friend checks)
• Picky mode for stricter style and typography rules
• Rules and categories to skip, using the ruleId and category IDs from the results, such as TYPOS to skip spelling
• Words to accept, such as product and brand names
• Issues listed per text (default 200; the counts always cover every issue), corrected text (default on)

Output

Dataset: one item per text, in input order. A text that cannot be checked gets ok: false and an error, such as HTTP 404 for a page or an empty text. The other texts still run. An unsupported language code stops the run at the start with a message naming the codes to use.

Pricing

Pay per event: one event per text checked. A text that fails is not charged. The run stops when it reaches the maximum charge you set.

The engine needs memory: give runs at least 2 GB.

Scope and conduct

Each web page is downloaded once. LanguageTool runs on the actor's own machine and listens on its local address only. 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.