Streaming Catalog Metadata Harmonization Framework

Role: Senior Media Catalog Architect specializing in video metadata engineering and content discovery systems.

Context

• Target Platform: {{streaming_platform_name}}
• Ingestion Feeds: {{catalog_source_types}}
• Baseline Taxonomy Standard: {{target_schema_standard}}
• Entity Match Criteria: {{duplicate_threshold_rules}}
• Global Distribution Targets: {{localization_locales}}
• Data Quality SLA: {{taxonomy_governance_tier}}

Task

Design a comprehensive data cleaning framework that systematically audits, normalizes, deduplicates, and validates incoming video asset records from {{catalog_source_types}} into {{target_schema_standard}} to maximize search discoverability and recommendation precision on {{streaming_platform_name}}.

Method

1. Map every inbound raw field from {{catalog_source_types}} against {{target_schema_standard}}, identifying critical structural discrepancies, type mismatches, and missing mandatory keys.
2. Define regex and programmatic sanitization routines for core title strings, episode naming sequences, rating designations, and runtime representations.
3. Formulate deterministic and probabilistic entity resolution algorithms using {{duplicate_threshold_rules}} to detect identical titles across varying distribution windows.
4. Construct localized metadata reconciliation pipelines across {{localization_locales}}, resolving character encoding corruption, language tag mismatches, and regional title variants.
5. Design categorical hierarchy transformation logic to reconcile conflicting genre, sub-genre, and mood tags into an authoritative controlled vocabulary.
6. Specify automated credit, cast, and crew attribution parsing to standardize name variants, role codes, and order billing.
7. Build a multi-tier exception routing protocol aligned with {{taxonomy_governance_tier}} for quarantined records that fail automated cleaning.
8. Establish automated validation gates and schema drift monitors to catch corrupted provider payloads before ingestion.

Constraints

• MUST preserve immutable raw partner IDs alongside unified internal master media identifiers.
• MUST NOT drop partial records if core visual identifiers (EIDR, TMS/Gracenote ID, or internal SKU) can be verified.
• All cleaning rules MUST prioritize deterministic matching before applying fuzzy matching thresholds.
• Transformation logic MUST specify exact fallback values for missing non-critical attributes.

Output format

Provide the operational framework using these exact sections:

1. Metadata Ingest & Profiling Baseline (bulleted field audit table)
2. String Sanitization & Normalization Protocol (step-by-step transformation rules)
3. Deduplication & Entity Resolution Engine (logic definition with threshold formulas)
4. Internationalization & Localization Cleansing Matrix (locale-specific handling rules)
5. Exception Handling & Governance Runbook (quarantine flow and remediation triggers) Total response length must be between 600 and 900 words.

Self-review

• Did I define concrete normalization logic for both structured attributes and localized text strings?
• Are all variable inputs ({{streaming_platform_name}}, {{catalog_source_types}}, {{target_schema_standard}}, {{duplicate_threshold_rules}}, {{localization_locales}}, {{taxonomy_governance_tier}}) actively referenced in the framework?
• Does the quarantine mechanism align strictly with the stated governance tier?