Attribute Enrichment
Attribute enrichment automatically extracts and enhances product attributes from your catalog data using LLM-based ranking and relevance models, improving searchability and discoverability without manual work.
Overview
Marqo's LLM Based Ranking and Relevance system automatically enriches product attributes as part of its ranking optimization process. The system extracts attributes from product descriptions, titles, and other unstructured data, normalizes values for consistency, and enhances product information to improve search relevance.
Benefits
- Automatic extraction: Attributes are extracted automatically from your existing product data—no manual tagging required
- Improved search relevance: Enriched attributes help products match more accurately to customer queries
- Better facet quality: Normalized attributes create cleaner, more useful facets for filtering
- Consistent data: Automatic normalization ensures consistent attribute values across your catalog
- Zero maintenance: The system continuously improves attribute quality as it learns from customer interactions
How It Works
The LLM Based Ranking and Relevance system analyzes your product catalog and:
- Extracts attributes from product descriptions, titles, and specifications
- Normalizes values to ensure consistency (e.g., "navy", "dark blue", "midnight" → "navy")
- Enhances data by inferring missing attributes based on product context
- Continuously improves as it learns from customer search and purchase patterns
This happens automatically in the background—no configuration or manual work required.
Accessing Enriched Attributes
If you have Marqo Enterprise enabled, you can retrieve the enriched attributes via the "Get Attributes" API. This allows you to:
- Export enriched attributes for use in other systems
- Review and validate attribute quality
- Integrate enriched data into your product management workflows
Get Attributes API Example
Retrieve enriched attributes for a specific product:
curl -X GET "https://ecom.marqo-ep.ai/api/v1/indexes/{index_name}/attributes/{product_id}" \
-H "x-marqo-index-id: {index_id}" \
-H "Content-Type: application/json"
Response:
{
"productId": "prod_123",
"enrichedAttributes": {
"color": "navy",
"material": "cotton",
"style": "casual",
"size": "large",
"pattern": "solid",
"occasion": "everyday",
"tags": ["comfortable", "breathable", "versatile"]
},
"originalAttributes": {
"color": "dark blue",
"material": "100% cotton",
"size": "L"
},
"confidenceScores": {
"color": 0.95,
"material": 0.98,
"style": 0.87,
"size": 1.0,
"pattern": 0.92,
"occasion": 0.85
}
}
You can also retrieve enriched attributes for multiple products or filter by specific attributes using query parameters.
Related Topics
- LLM Based Ranking and Relevance - Learn how the ranking system works
- Synonyms & Taxonomy - Manage attribute synonyms