Scraped Text Sanitizer

Scraped Text Sanitizer repairs dirty, garbled text in records you already have — no scraping, no login, no API keys. Paste an array of records (or a single string) and it fixes mojibake, decodes HTML entities, strips leftover HTML tags, applies Unicode NFKC normalization, optionally removes emoji, and collapses whitespace — every step individually toggleable. You get back clean text plus a per-record summary of exactly what changed.

It runs on the Apify platform, so you get API access, scheduling, integrations (Make, Zapier, n8n), and pay-per-record billing with no infrastructure to manage.

What does Scraped Text Sanitizer do?

Text harvested from the web, PDFs, CSV exports, or old databases is frequently broken: Café instead of Café, Tom & Jerry instead of Tom & Jerry, stray <p>/<span> tags, fullwidth and ligature characters, and messy whitespace. This Actor cleans all of that in one pass. It only transforms data you supply — it never fetches, crawls, or logs into anything.

Under the hood it uses the battle-tested ftfy library for mojibake repair, Python's standard html and unicodedata for entity decoding and NFKC normalization, and deterministic rules for tag stripping and whitespace collapsing.

Why use Scraped Text Sanitizer?

• Fix mojibake at scale — repair é, ’, and other encoding damage without writing a script.
• Clean up scraper output — decode entities and strip HTML tags that leak into extracted text.
• Normalize for search & dedup — NFKC folds fullwidth forms and ligatures so records match.
• Full transparency — every record reports which operations changed it, so you can trust the result.
• No code, no servers — run on demand, on a schedule, or from your own app via the Apify API.

How to use Scraped Text Sanitizer

1. Open the Actor and go to the Input tab.
2. Paste your records into the Records to clean field (an array of strings, or objects with a text field). Optionally add a single text string.
3. Toggle the cleaning operations you want (all on by default except Strip emoji).
4. Click Start. When it finishes, open the Output tab or download the dataset as JSON, CSV, Excel, or HTML.

Input

Field	Type	Description
records	array	Records to clean. Each item is a string, or an object containing the text field named by textField. Add an id to join results back to your source.
text	string	Optional single string to clean, processed in addition to records.
textField	string	For object records, the property holding the text (default text).
fixMojibake	boolean	Repair encoding damage / mojibake (default true).
decodeHtmlEntities	boolean	&amp; → &, &lt; → < (default true).
stripHtmlTags	boolean	Remove leftover HTML tags (default true).
normalizeUnicode	boolean	Unicode NFKC normalization (default true).
stripEmoji	boolean	Remove emoji and pictographs (default false).
collapseWhitespace	boolean	Collapse whitespace runs to single spaces and trim (default true).

Input example

{
  "records": [
    { "id": "a", "text": "Caf&eacute; &amp; Crêpes <b>Paris</b>   " },
    { "id": "b", "text": "Rock &amp; Röll finale" }
  ],
  "stripEmoji": false
}

Output

Each input record produces one dataset item with the cleaned text and a change summary. You can download the dataset in various formats such as JSON, HTML, CSV, or Excel.

{
  "id": "a",
  "recordIndex": 0,
  "cleanedText": "Café & Crêpes Paris",
  "originalText": "Caf&eacute; &amp; Crêpes <b>Paris</b>   ",
  "changed": true,
  "changes": {
    "fixedMojibake": true,
    "decodedHtmlEntities": true,
    "strippedHtmlTags": true,
    "normalizedUnicode": false,
    "collapsedWhitespace": true
  },
  "originalLength": 41,
  "cleanedLength": 19,
  "billed": true
}

Output fields

Field	Type	Description
id	string/number	Pass-through identifier from the input record, if supplied.
recordIndex	integer	Zero-based position of the record in the processed input.
cleanedText	string	The sanitized text after all enabled operations.
originalText	string	The exact input text before cleaning.
changed	boolean	true if the cleaned text differs from the original.
changes	object	Per-operation flags — true if that enabled operation modified the text. Only enabled operations appear.
originalLength	integer	Character count of the original text.
cleanedLength	integer	Character count of the cleaned text.
billed	boolean	true if the record was charged as a paid result (non-empty output under pay-per-event).

Pricing

This Actor uses pay-per-result billing: you are charged once per non-empty cleaned record. Empty or whitespace-only inputs are returned for transparency but are never charged. Compute is negligible (tens of thousands of records per second), so cost is dominated by the per-record price set in the Console. See PRICING.md for the unit-economics rationale.

Tips

• Leave Strip emoji off unless you specifically need emoji removed — they are often meaningful content.
• Disable operations you don't need (e.g. keep HTML tags) via the toggles; the change summary reflects only the operations you enabled.
• Use the id field on records to join cleaned output back to your source rows.

FAQ and support

• Does it scrape or fetch anything? No. It only transforms data you pass in as input. There is no network access, authentication, or PII harvesting.
• What fixes mojibake? The ftfy library's encoding repair, applied before entity decoding and normalization.
• Something looks wrong? Open the Issues tab with an example input and the expected output.