Review Dataset Auditor

Audit heterogeneous review records before analytics, triage, or response workflows. The Actor produces one deterministic report covering usable review text, rating/date/author completeness, duplicate IDs, and exact normalized-text duplicates across common Google Maps, Amazon, App Store, and generic review fields.

No external API key or LLM is required. The Actor makes no external model or third-party API calls.

Input

Use the Source Dataset READ-only resource picker, pass reviews inline, or run {} for a clearly labeled embedded sample. Input precedence is: presence of reviews (including []), then nonblank datasetId, then sample.

{
  "datasetLabel": "May review export",
  "reviews": [
    { "id": "r-1", "text": "Exports fail every afternoon.", "rating": 2, "date": "2026-05-01" },
    { "id": "r-1", "text": "Exports fail every afternoon.", "rating": 2, "date": "2026-05-01" }
  ],
  "maxDuplicateGroups": 25
}

Output and pricing shape

Every run writes exactly one review_dataset_audit_report Dataset item and the same report to OUTPUT. One aggregate audit equals one Dataset result. Synthetic Dataset-item charging is automatic; the code contains no manual charging.

The quality score is a bounded deterministic readiness indicator, not a fraud or authenticity score. Exact duplicate content is matched only after lowercasing and whitespace normalization; semantic near-duplicates are not inferred.

Limits and privacy

• At most 5,000 source records and 100 returned duplicate groups.
• Review text is inspected through the tested Review Intelligence normalization boundary; no raw review body is copied into duplicate-group output.
• Source Dataset access is READ-only and never writes to or deletes the source.
• User data is written only to the run's default Dataset and OUTPUT; Apify retention settings apply.