Vocabulary-Aware Search Typo Corrector

Suggest conservative corrections to search queries using your own vocabulary and dictionary. Keep original queries, protected words and ambiguous matches intact.

Run and download

1. Keep the included sample or replace Input data with your own authorized records.
2. Start a run. Open Report and CSV downloads when it finishes.
3. Open OUTPUT for the complete JSON report and download the CSV records.
4. Review the findings before importing or publishing changes. This Actor never writes to a store.

Hosted limits

Up to 50 queries, 5000 supplied corpus words, 10000 dictionary entries, 1000 protected entries or 20 documents of 10000 characters each; 1 MiB total. These tighter hosted limits take precedence over the core format limits below. Memory is fixed at 256 MB. Default timeout is 60 seconds. Output includes one dataset summary row; individual findings and records are delivered in OUTPUT and CSV files. No model subscription, external API key or other product purchase is required.

Price

$0.05 per run, charged once when the Actor starts. All report and CSV output is included, with no per-row fee or platform-usage surcharge. Set the maximum run cost to at least $0.05. Starts that reject invalid input or produce no findings still incur the start charge; validate the input before running.

Storage and support

Apify stores supplied data and generated reports under its storage access/retention settings. This hosted edition sends input to Apify; it is not the local-only edition. Remove sensitive run data when no longer needed. URLs in input are never fetched. Use the Actor issue tracker for support with a non-sensitive example and run ID. Do not post private exports or credentials. Hosted use does not grant source redistribution rights.

Input contains either corpus {word: frequency} or documents array, plus dictionary, optional protected words, queries, and optional dominanceRatio (default 3, minimum 2). All fields are demonstrated in the prefilled example. A real general-language dictionary is required for deployment; the tiny sample list demonstrates behavior only. No dictionary or third-party dictionary rights are sold with this product. Use a list you have rights to use and retain its notices when redistributing your data.

Words already in the corpus, dictionary or protected list stay unchanged. Other words need a unique edit-distance-one candidate or the configured frequency advantage. Ambiguous results explicitly remain unchanged. Mixed alphanumeric identifiers are not corrected. Original queries are never overwritten; use retrievalQuery for search only and offer users the unchanged original as an override. Full JSON includes individual refusal decisions, corpus/dictionary hashes and metadata. JSON is authoritative; CSV guards formula-like prefixes.

Limits: 1 MiB input, 5000 supplied corpus words, 10000 dictionary words, 1000 protected words, 50 queries of 500 characters; words at most 100 characters. Documents: at most 20 of 10000 characters each. Counts are positive integers at most 1 billion. Length buckets reduce candidate search. Unicode letters supported; this is not a language model or proof of intended meaning. Dictionary quality and corpus freshness determine behavior.